{ "edition": "2026-10-24", "published_at": "2026-10-24T16:52:34Z", "rubric_version": "2.0", "models": { "jev": "jev-1.14.0", "judge_a": "claude-haiku-5-5", "judge_b": "gpt-5.6-luna", "writer": "claude-sonnet-5-5", "summarizer": "claude-haiku-5-5", "claim_check": "claude-sonnet-5-5 + jev-1.14.0" }, "stats": { "window_start": "2026-10-23T00:00:00.000Z", "window_end": "2026-10-24T00:00:00.000Z", "candidates": 1627, "clusters": 1552, "scored": 1552, "dropped_off_topic": 208, "stage_b": 243, "claim_checked": 20, "cost_usd": 6.8321 }, "feed": { "score_floor": 0.55, "shown_first": 20, "page_size": 20, "default_hidden": [], "default_hidden_formats": [ "new_projects" ] }, "categories": [ { "key": "programming", "name": "Programming & dev tools", "description": "Programming and developer tools (languages, frameworks, dev tools, engineering practice)", "count": 88 }, { "key": "ai", "name": "AI & machine learning", "description": "AI and machine learning (models, applied AI, AI tooling)", "count": 180 }, { "key": "security", "name": "Security", "description": "Security (vulnerabilities, incidents, privacy engineering)", "count": 63 }, { "key": "devops", "name": "Infrastructure & DevOps", "description": "Infrastructure and DevOps (cloud, SRE, platforms, CI/CD, Kubernetes)", "count": 37 }, { "key": "it", "name": "IT & networking", "description": "IT and networking (enterprise IT, sysadmin, networks, end-user computing)", "count": 13 }, { "key": "data", "name": "Data", "description": "Data (databases, data engineering, analytics)", "count": 14 }, { "key": "hardware", "name": "Hardware & devices", "description": "Hardware and devices (chips, embedded, devices)", "count": 15 }, { "key": "product", "name": "Product", "description": "Product management (discovery, roadmaps, pricing, product strategy)", "count": 3 }, { "key": "design", "name": "Design & UX", "description": "Design and UX (interface and product design, user research)", "count": 3 }, { "key": "growth", "name": "Growth & marketing", "description": "Growth and marketing (acquisition, positioning, content)", "count": 1 }, { "key": "startups", "name": "Startups & founders", "description": "Startups and founders (company building, fundraising, tech business)", "count": 1 }, { "key": "fintech", "name": "Fintech", "description": "Fintech (payments, banking technology, crypto infrastructure)", "count": 2 }, { "key": "policy", "name": "Policy & regulation", "description": "Policy and regulation (privacy law, AI rules, antitrust)", "count": 11 }, { "key": "other", "name": "Other", "description": "Stories that fit none of the topics", "count": 1 } ], "formats": [ { "key": "original_research", "name": "Papers & studies", "description": "Original research (a paper, study or new measurement)", "count": 128 }, { "key": "engineering_writeup", "name": "Engineering write-ups", "description": "Engineering write-up (how a team built, ran or fixed something)", "count": 33 }, { "key": "release", "name": "Releases", "description": "Release (a product, model, library or version announcement)", "count": 25 }, { "key": "analysis_opinion", "name": "Analysis & opinion", "description": "Analysis or opinion", "count": 20 }, { "key": "tutorial", "name": "Tutorials", "description": "Tutorial or how-to", "count": 2 }, { "key": "news_report", "name": "News", "description": "News report about events or other people's work", "count": 27 }, { "key": "other", "name": "Other", "description": "Other", "count": 3 }, { "key": "new_projects", "name": "New projects", "description": "New side projects and young GitHub repos with few stars, shared by their makers", "count": 7 } ], "stories": [ { "rank": 1, "id": "st_20261024_1312", "title": "The attack chain behind the v1.107.0 LiteLLM release", "url": "https://docs.litellm.ai/blog/security-incident-oct-21-2026-attack-chain", "author": null, "site": "docs.litellm.ai", "published_at": "2026-10-23T21:05:16.000Z", "reading_minutes": 10, "depth": "claim_checked", "format": "engineering_writeup", "summary": "LiteLLM's root-cause analysis says the malicious 1.107.0 release came from a master key committed to a public repo on May 27 and never revoked. On October 10 it yielded an org owner's GitHub token, which the attacker used to harvest two maintainers' tokens through CI. On October 21 they repointed the release workflow to a fork and approved the release with those tokens. The relaxed two-engineer approval rule helped. These details come from LiteLLM alone and were not independently verified.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "programming", "ai", "security", "devops" ] }, "score": { "total": 0.8332, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.708, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7558, "spread": 0.16, "weight": 0.25 }, "reader_interest": { "jev": 0.86, "judge_a": 0.92, "judge_b": 0.97, "mean": 0.9167, "spread": 0.11, "weight": 0.2 }, "audience_breadth": { "jev": 0.33, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.65, "spread": 0.49, "weight": 0.1 }, "practical_relevance": { "jev": 0.85, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8667, "spread": 0.05, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9444, "spread": 0.08, "weight": 0.1 }, "technical_depth": { "jev": 0.968, "judge_a": 0.9, "judge_b": 0.94, "mean": 0.9358, "spread": 0.068, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.95, "judge_b": 0.99, "mean": 0.9533, "spread": 0.07, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.13, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.0867, "spread": 0.1, "weight": -0.15 } } }, "claims": [ { "text": "On May 27, 2026, a configuration file containing the master key to a LiteLLM Agent Platform deployment was committed to the public LiteLLM-Labs/lite-harness repository.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "On May 27, 2026, a configuration file was committed to the public LiteLLM-Labs/lite-harness repository which inadvertently contained the master key", "source_quote": null, "source_url": "https://github.com/LiteLLM-Labs/lite-harness/blob/e10b14a/harnesses/opencode/opencode.json" }, { "text": "On October 10 the threat actor used the exposed master key to access the LiteLLM Agent Platform and extract a GitHub CLI OAuth token belonging to the GitHub organization owner, giving owner-level access to all repositories.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "On October 10th, the threat actor used this key to gain access to our LiteLLM Agent Platform and extracted this GitHub CLI OAuth token", "source_quote": null, "source_url": null }, { "text": "The malicious releases were published to LiteLLM's public registries at around 4:40 AM PDT (11:40 UTC) on October 21.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "The malicious releases were published to our public registries at around 4:40 AM PDT (11:40 UTC", "source_quote": null, "source_url": null }, { "text": "The PyPI releases were frozen by 6:23 AM PDT and the Docker images were removed by 7:20 AM PDT.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "The PyPi releases were frozen by 6:23 AM PDT (13:23 UTC", "source_quote": null, "source_url": null }, { "text": "The release restriction requiring approval by a special account locked to a physical device was relaxed earlier that week, allowing any two engineers to trigger and approve a release.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "This restriction was relaxed earlier this week to accommodate the temporary push for more frequent releases and bugfix backports, allowing any two engineers to trigger and approve a release", "source_quote": null, "source_url": null }, { "text": "Seven litellm==1.107.0 wheels were published to PyPI between 11:39:51 and 11:40:13 UTC on October 21.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "Seven PyPI wheels went out between 11:39:51 and 11:40:13", "source_quote": null, "source_url": null }, { "text": "LidorMachluf caught the malicious release within 20 minutes of publication.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "catching the malicious release within 20 minutes of publication", "source_quote": null, "source_url": "https://github.com/BerriAI/litellm/issues/47826" }, { "text": "The compromised artifacts were live for about two hours before being pulled.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "The compromised artifacts were live for about two hours before they were pulled", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 17, "front_page": true, "url": "https://news.ycombinator.com/item?id=50215317" }, "lobsters": null, "line": "Elsewhere: HN 17 pts, front page", "checked_at": "2026-10-24T08:53:56.860Z" } }, { "rank": 2, "id": "st_20261024_153", "title": "eMudhra certificates are compromised. Here's what you need to know", "url": "https://blog.cloudflare.com/emudhra-certificate-compromise/", "author": null, "site": "Cloudflare Blog", "published_at": "2026-10-23T04:28:28.000Z", "reading_minutes": 4, "depth": "claim_checked", "format": "engineering_writeup", "summary": "Private keys for eight eMudhra intermediate CA certificates, including one reportedly equal to the integer 4, are weak enough that anyone could forge certificates for any domain. Cloudflare says it patched its clients, Tunnel and CDN, and found no exploitation on its services, though these details are Cloudflare's own and unverified. Customers must still update origin and mutual TLS trust stores, container images and dependencies, and not disable verification. Browsers can revoke quickly, but other software needs per-stack updates.", "also_covered_by": [ { "site": "Unmitigated Risk", "url": "https://unmitigatedrisk.com/2026/10/trusted-everywhere-impact-everywhere/", "title": "Trusted Everywhere, Impact Everywhere" } ], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "security", "devops", "it" ] }, "score": { "total": 0.8097, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.798, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8058, "spread": 0.022, "weight": 0.25 }, "reader_interest": { "jev": 0.78, "judge_a": 0.92, "judge_b": 0.95, "mean": 0.8833, "spread": 0.17, "weight": 0.2 }, "audience_breadth": { "jev": 0.937, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8722, "spread": 0.137, "weight": 0.1 }, "practical_relevance": { "jev": 0.99, "judge_a": 0.92, "judge_b": 0.94, "mean": 0.95, "spread": 0.07, "weight": 0.15 }, "substantiated": { "jev": 0.637, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7122, "spread": 0.3, "weight": 0.1 }, "technical_depth": { "jev": 0.81, "judge_a": 0.75, "judge_b": 0.78, "mean": 0.78, "spread": 0.06, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9367, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.15, "judge_b": 0.03, "mean": 0.08, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.34, "judge_a": 0.2, "judge_b": 0.04, "mean": 0.1933, "spread": 0.3, "weight": -0.15 } } }, "claims": [ { "text": "Cloudflare became aware on October 19, 2026 at approximately 17:00 UTC that private keys for eight eMudhra intermediate certificates were compromised.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "On October 19, 2026, at approximately 17:00 UTC, Cloudflare became aware that private keys associated with eight eMudhra intermediate certificates were compromised", "source_quote": null, "source_url": null }, { "text": "Because the keys are weak, an attacker can create a forged certificate for any domain and impersonate servers to intercept or modify traffic.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "Because the keys themselves are weak, an attacker can create a forged certificate for any domain and use it to impersonate servers to intercept or modify traffic", "source_quote": null, "source_url": null }, { "text": "Cloudflare's security review has not identified active exploitation of this issue on Cloudflare services.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "Our security review has not identified active exploitation of this issue on Cloudflare services", "source_quote": null, "source_url": null }, { "text": "Cloudflare Tunnel (cloudflared) version 2026.10.2 is the upgrade that addresses the issue.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Cloudflare Tunnel (cloudflared) Upgrade to 2026.10.2", "source_quote": null, "source_url": "https://github.com/cloudflare/cloudflared/releases" }, { "text": "Cloudflare One Client fixed versions are v2.5.7 (Android), v1.12.7 (iOS), and 2026.8.2115.0 on Windows, macOS and Linux.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Upgrade to: v2.5.7 (Android), v1.12.7 (iOS), Windows Desktop (2026.8.2115.0", "source_quote": null, "source_url": "https://downloads.cloudflareclient.com/v1/download/windows/version/2026.8.2115.0" }, { "text": "Cloudflare's CDN and server-side Zero Trust mitigations were fully deployed on October 20 at 19:20 UTC.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "October 20 19:20 UTC CDN and server-side Zero Trust mitigations fully deployed", "source_quote": null, "source_url": null }, { "text": "Mozilla's Dennis Jackson has crowdsourced a list of commonly used packages that have been patched.", "kind": "quote", "label": "matches", "drift": null, "article_quote": "Mozilla’s Dennis Jackson has crowdsourced a list of commonly used packages that have been patched", "source_quote": "Site published by Dennis Jackson . This site is the work of many contributors from the community. The source is on GitHub and PRs are welcome", "source_url": "https://emudhra.we-have-a-problem.com/" }, { "text": "Browser maintainers could remediate the issue quickly because most browsers use quick-revocation mechanisms such as Chrome CRLSets.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "Browser maintainers were able to remediate the issue with a flip of a switch", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204516" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:56.889Z" } }, { "rank": 3, "id": "st_20261024_90", "title": "Full analysis of the .ida \"Code Red\" worm", "url": "https://seclists.org/bugtraq/2001/Jul/359", "author": "Marc Maiffret", "site": "seclists.org", "published_at": "2026-10-23T02:28:42.000Z", "reading_minutes": 18, "depth": "claim_checked", "format": "engineering_writeup", "summary": "eEye Digital Security published a disassembly-based analysis of the \"Code Red\" worm, which spreads through the IIS .ida buffer overflow patched in Microsoft bulletin MS01-033. Each infected server runs 100 threads that probe addresses generated from a static seed, so hosts rescan the same sequence and generate heavy traffic. Time-gated floods against www.whitehouse.gov and a defacement on US English systems follow. eEye's figures, including at least 12,000 infections, come from its own testing and administrator reports, not independent verification. For builders, the point is patch speed.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "security", "it" ] }, "score": { "total": 0.8017, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.8, "judge_a": 0.85, "judge_b": 0.88, "mean": 0.8433, "spread": 0.08, "weight": 0.25 }, "reader_interest": { "jev": 0.41, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.7633, "spread": 0.57, "weight": 0.2 }, "audience_breadth": { "jev": 0.723, "judge_a": 0.85, "judge_b": 0.72, "mean": 0.7644, "spread": 0.13, "weight": 0.1 }, "practical_relevance": { "jev": 0.733, "judge_a": 0.9, "judge_b": 0.86, "mean": 0.8308, "spread": 0.167, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8989, "spread": 0.097, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.9, "judge_b": 0.96, "mean": 0.9492, "spread": 0.088, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9433, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.0767, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.2033, "spread": 0.07, "weight": -0.15 } } }, "claims": [ { "text": "eEye Digital Security discovered the .ida vulnerability exploited by the worm and published an advisory on it on June 18, 2001.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "the recent .ida vulnerability that eEye Digital Security discovered", "source_quote": null, "source_url": "http://www.eeye.com/html/Research/Advisories/AD20010618.html" }, { "text": "The Code Red worm's ultimate purpose appears to be a denial of service attack against www.whitehouse.gov.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "this worms purpose ultimately seems to be to perform a denial of service attack against www.whitehouse.gov", "source_quote": null, "source_url": null }, { "text": "Only US English Windows NT/2000 systems show the defaced 'Hacked by Chinese!' web page.", "kind": "generalization", "label": "no_source", "drift": null, "article_quote": "only US English Windows NT/2000 systems will show the defaced (\"Hacked by Chinese !\") web page", "source_quote": null, "source_url": null }, { "text": "eEye's testing calculated that the worm can attempt to infect roughly half a million IP addresses a day.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "the worm can attempt to infect roughly half a million IP addresses a day", "source_quote": null, "source_url": null }, { "text": "Administrators reported being probed by over 12 thousand unique hosts, which eEye took to mean at least 12 thousand hosts were infected.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "administrators that have been probed by over 12 thousand unique hosts. That basically means at least 12 thousand hosts have been infected by this worm", "source_quote": null, "source_url": null }, { "text": "Every infected computer goes through the same list of 'random' IP addresses because the worm uses a static seed, causing reinfection and traffic that creates a denial-of-service effect.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "every computer infected by this worm is going to go through the same list of random IP addresses to try to infect", "source_quote": null, "source_url": null }, { "text": "The worm exploits the .ida buffer overflow by overwriting EIP with 0x7801CBD3, an address in msvcrt.dll containing a 'call ebx' instruction.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "EIP is overwritten with 0x7801CBD3 which an address within msvcrt.dll. The code at 0x7801CBD3 disassembles to", "source_quote": null, "source_url": null }, { "text": "A Microsoft patch for the .ida vulnerability (MS01-033) had been available for over a month before the worm spread.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "A patch for this vulnerability has been available for over a month now", "source_quote": null, "source_url": "http://www.microsoft.com/technet/treeview/default.asp?url=/technet/security" } ], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203807" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:56.875Z" } }, { "rank": 4, "id": "st_20261024_661", "title": "eMudhra ignored Mozilla Root investigation deadline; blames social media", "url": "https://bugzilla.mozilla.org/show_bug.cgi?id=2083449", "author": "Wayne", "site": "bugzilla.mozilla.org", "published_at": "2026-10-23T14:52:08.000Z", "reading_minutes": 35, "depth": "claim_checked", "format": "news_report", "summary": "Eight eMudhra intermediate CAs were revoked for caCompromise after outsiders found their P-384 private keys were tiny integers (below 256), recoverable almost instantly. Six issued TLS certificates chaining to roots trusted by major browsers. eMudhra attributes the flaw to a key-ceremony script error, and Sectigo revoked the cross-signs. Bun 1.4.4 now removes the root and blocks the eight keys, since it checks revocation only via a user-supplied CRL. eMudhra announced its exit from global TLS. Teams with bundled trust stores should check them; several figures here come from unsourced claims.", "also_covered_by": [ { "site": "GitHub", "url": "https://github.com/oven-sh/bun/security/advisories/GHSA-26q8-h657-pf84", "title": "Bun 1.4.4 removes compromised eMudhra certs" }, { "site": "SSLBoard", "url": "https://sslboard.com/blog/emudhra-revoked-certificate.md/", "title": "eMudhra served a revoked certificate for four days, then switched to Sectigo" } ], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security" ] }, "score": { "total": 0.7613, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.778, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8258, "spread": 0.123, "weight": 0.25 }, "reader_interest": { "jev": 0.8, "judge_a": 0.85, "judge_b": 0.96, "mean": 0.87, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.777, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.7322, "spread": 0.077, "weight": 0.1 }, "practical_relevance": { "jev": 0.755, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.7783, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.94, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.91, "spread": 0.09, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.9067, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9267, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.17, "judge_a": 0.02, "judge_b": 0.01, "mean": 0.0667, "spread": 0.16, "weight": -0.15 }, "overclaim": { "jev": 0.65, "judge_a": 0.2, "judge_b": 0.62, "mean": 0.49, "spread": 0.45, "weight": -0.15 } } }, "claims": [ { "text": "The private keys of 8 eMudhra intermediate certificates were reported compromised to Mozilla on 17 October, and Mozilla confirmed the proof of key control was valid.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "On 17th October @ 20:06 UTC, Mozilla received a report to certificates@mozilla.org, forwarded by the Chrome Root program, from Wayne, that the private key of 8 eMudhra Intermediate certificates have…", "source_quote": null, "source_url": null }, { "text": "The 8 impacted P384 keys all used private key integers less than 256, so they can be recovered near instantly.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "The 8 impacted keys all used integers less than 256. That is, the private keys are effectively all-zeroes apart from the last byte", "source_quote": null, "source_url": null }, { "text": "Mozilla completed distrust of the impacted certificates in Firefox via OneCRL at 22:59 UTC on 17 October.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "This process was tracked in Bug 2083382 and completed @ 22:59 UTC", "source_quote": null, "source_url": "https://bugzilla.mozilla.org/show_bug.cgi?id=2083382" }, { "text": "eMudhra said the weak keys resulted from a flaw in the key generation ceremony script during an approved wrap/unwrap test deviation, and the weaker second key pair was used by mistake because key labels were similar.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "a flaw in the code resulted in the second pair of keys being created onboard which were weaker", "source_quote": null, "source_url": null }, { "text": "Mozilla decided on emergency removal of trust in emSign ECC Root CA - G3 after multiple individuals independently recovered the private keys.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "we have decided that the circumstances warrant the emergency removal of trust in emSign ECC Root CA - G3", "source_quote": null, "source_url": null }, { "text": "eMudhra estimated about 30,000 certificates need migration, about 97% issued via ACME.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "there are about 30,000 certs that require migration and about 97% came via ACME", "source_quote": null, "source_url": null }, { "text": "Mozilla sent questions to eMudhra on 20 October asking for a response within 24 hours; after an extension request, no response had been received by Friday 23rd.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "As of the time of writing on Friday 23rd, Mozilla has not received a response to these questions", "source_quote": null, "source_url": null }, { "text": "eMudhra said it recently voluntarily decided to withdraw from the global SSL/TLS certificate business.", "kind": "quote", "label": "matches", "drift": null, "article_quote": "decided to voluntarily withdraw from the global SSL/TLS certificate business", "source_quote": "its Board has decided to voluntarily withdraw from the global SSL/TLS certificate business", "source_url": "https://emudhra.com/en/press-release/emudhra-to-exit-global-ssl-tls-certificate-business" } ], "elsewhere": { "hn": { "points": 201, "front_page": true, "url": "https://news.ycombinator.com/item?id=50210067" }, "lobsters": null, "line": "Elsewhere: HN 201 pts, front page", "checked_at": "2026-10-24T08:53:56.895Z" } }, { "rank": 5, "id": "st_20261024_733", "title": "Ruby RegEx could lead to RCE", "url": "https://www.ruby-lang.org/en/news/2026/10/23/regexp-fixes-for-gitlab-cves/", "author": null, "site": "Ruby Programming Language", "published_at": "2026-10-23T15:55:02.000Z", "first_published_at": "2026-10-23T00:00:00.000Z", "reading_minutes": 3, "depth": "claim_checked", "format": "release", "summary": "Ruby 4.0.9, 3.4.12 and 3.3.13 fix regular expression engine (Onigmo) bugs that GitLab reported could give remote code execution when an application compiles an attacker-supplied pattern, for example via Regexp.new or /#{input}/. GitLab issued four CVEs; Ruby has requested none yet. Regexp.timeout and length limits do not reliably mitigate them. Apps using only developer-written patterns are unaffected. Most claims here come from Ruby's own notice and were not independently verified.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "security" ] }, "score": { "total": 0.7582, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.648, "judge_a": 0.55, "judge_b": 0.8, "mean": 0.6658, "spread": 0.25, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.72, "judge_b": 0.95, "mean": 0.8133, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.663, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.6278, "spread": 0.063, "weight": 0.1 }, "practical_relevance": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.88, "spread": 0.14, "weight": 0.15 }, "substantiated": { "jev": 0.93, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.92, "spread": 0.13, "weight": 0.1 }, "technical_depth": { "jev": 0.79, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7233, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.94, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0233, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.1367, "spread": 0.15, "weight": -0.15 } } }, "claims": [ { "text": "Ruby released versions 4.0.9, 3.4.12 and 3.3.13, which fix bugs in Ruby's regular expression engine (Onigmo) that GitLab found could be used for remote code execution on GitLab servers.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "We have released Ruby 4.0.9, 3.4.12 and 3.3.13. They fix several bugs in Ruby’s regular expression engine (Onigmo) that GitLab found could be used for remote code execution on GitLab…", "source_quote": null, "source_url": null }, { "text": "CVE-2026-89078 is a double free issue in the Regular Expression Parser, fixed in GitLab 19.4.1.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Double Free issue in Regular Expression Parser", "source_quote": null, "source_url": "https://www.cve.org/CVERecord?id=CVE-2026-89078" }, { "text": "CVE-2026-97229 is an integer overflow issue in the CI/CD Regular Expression Compiler, fixed in GitLab 19.4.2.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Integer Overflow issue in CI/CD Regular Expression Compiler", "source_quote": null, "source_url": "https://www.cve.org/CVERecord?id=CVE-2026-97229" }, { "text": "CVE-2026-110264 is an integer overflow issue in the Regular Expression Compiler, fixed in GitLab 19.5.1.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Integer Overflow issue in Regular Expression Compiler (GitLab 19.5.1", "source_quote": null, "source_url": "https://www.cve.org/CVERecord?id=CVE-2026-110264" }, { "text": "The Ruby change #13332, fixing the double free when parsing a pattern fails after an alternation, first appears in Ruby 4.0.0, 3.4.12 and 3.3.13.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Double free when parsing a pattern fails after an alternation #13332 4.0.0 3.4.12 3.3.13", "source_quote": null, "source_url": "https://github.com/ruby/ruby/pull/13332" }, { "text": "Ruby 4.0.9, 3.4.12 and 3.3.13 contain all of the fixes for these CVEs.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "Ruby 4.0.9, 3.4.12 and 3.3.13 contain all of these fixes", "source_quote": null, "source_url": null }, { "text": "Regexp.timeout does not protect against these bugs, and limiting pattern length does not reliably protect either because some can be triggered by short patterns.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "Regexp.timeout does not protect against these bugs. Limiting the length of patterns does not reliably protect against them either, because some of them can be triggered by short patterns", "source_quote": null, "source_url": null }, { "text": "Ruby's advisory for CVE-2022-28738 treated a double free in Regexp compilation as a vulnerability after a comprehensive assessment.", "kind": "quote", "label": "matches", "drift": null, "article_quote": "We stated this in the advisory for CVE-2022-28738, where we nevertheless treated a double free in Regexp compilation as a vulnerability after a comprehensive assessment", "source_quote": "In this case, however, following a comprehensive assessment, we treat this issue as a vulnerability", "source_url": "https://www.ruby-lang.org/en/news/2022/04/12/double-free-in-regexp-compilation-cve-2022-28738" } ], "elsewhere": { "hn": { "points": 7, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211066" }, "lobsters": null, "line": "Elsewhere: HN 7 pts, not on front page", "checked_at": "2026-10-24T08:53:57.157Z" } }, { "rank": 6, "id": "st_20261024_6", "title": "Node2Shell: Node.js Brotli vulnerability allows remote code execution", "url": "https://www.aikido.dev/blog/node2shell-nodejs-brotli-rce-cve-2026-111497", "author": "Written bySooraj Shah", "site": "aikido.dev", "published_at": "2026-10-23T00:54:32.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "news_report", "summary": "Node.js released patches on October 21 (22.23.4, 24.22.1, 26.12.1) for CVE-2026-111497, nicknamed Node2Shell, an out-of-bounds write in its bundled Brotli decoder that can crash a process or allow code execution. Bun needs 1.4.4. The risk covers Express 5 request bodies, fetch() responses and direct node:zlib use. Researchers reported unreliable code execution on an unmodified 24.22.0 build, with no in-the-wild exploitation reported by October 22. End-of-life lines get no fix, and the Python Brotli and brotlicffi packages are also affected.", "also_covered_by": [ { "site": "nodejs.org", "url": "https://nodejs.org/en/blog/vulnerability/october-2026-security-releases", "title": "Node2Shell: From Brotli Data to Remote Code Execution" }, { "site": "opensourcemalware.com", "url": "https://opensourcemalware.com/blog/node2shell-affects-python", "title": "Node2Shell vuln affected Python too" } ], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "programming", "security", "devops" ] }, "score": { "total": 0.7464, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.74, "judge_a": 0.72, "judge_b": 0.82, "mean": 0.76, "spread": 0.1, "weight": 0.25 }, "reader_interest": { "jev": 0.85, "judge_a": 0.92, "judge_b": 0.94, "mean": 0.9033, "spread": 0.09, "weight": 0.2 }, "audience_breadth": { "jev": 0.697, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.7589, "spread": 0.103, "weight": 0.1 }, "practical_relevance": { "jev": 0.995, "judge_a": 0.95, "judge_b": 0.96, "mean": 0.9683, "spread": 0.045, "weight": 0.15 }, "substantiated": { "jev": 0.577, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.6856, "spread": 0.203, "weight": 0.1 }, "technical_depth": { "jev": 0.905, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.775, "spread": 0.205, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.91, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.35, "judge_b": 0.38, "mean": 0.2667, "spread": 0.31, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.2833, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 14, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203180" }, "lobsters": null, "line": "Elsewhere: HN 14 pts, not on front page", "checked_at": "2026-10-24T08:53:57.167Z" } }, { "rank": 7, "id": "st_20261024_766", "title": "Multi-Interpretable Tarballs", "url": "https://stoppels.ch/2026/10/22/multi-interpretable-tarballs.html", "author": null, "site": "stoppels.ch", "published_at": "2026-10-23T15:22:57.000Z", "first_published_at": "2026-10-22T00:00:00.000Z", "reading_minutes": 10, "depth": "three_judges", "format": "engineering_writeup", "summary": "A Spack developer shows that one tarball can extract to different file trees depending on the tool, with no error or warning. Comparing GNU tar 1.35 and Python's tarfile (3.12 to 3.14), he found disagreements over NUL-prefixed size fields, whitespace in pax records, repeated headers, trailing-slash directories and symlink overwrites. Python also misreads GNU incremental archives because of a magic-byte bug. About a quarter of nixpkgs and a third of Guix hash only the tarball. His advice is to record a tree hash after extraction as well.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "security" ] }, "score": { "total": 0.7443, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.608, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6125, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.84, "spread": 0.21, "weight": 0.2 }, "audience_breadth": { "jev": 0.723, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.6478, "spread": 0.123, "weight": 0.1 }, "practical_relevance": { "jev": 0.768, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.7692, "spread": 0.14, "weight": 0.15 }, "substantiated": { "jev": 0.84, "judge_a": 0.85, "judge_b": 0.91, "mean": 0.8667, "spread": 0.07, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.9, "judge_b": 0.9, "mean": 0.9317, "spread": 0.095, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9067, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.02, "judge_b": 0.01, "mean": 0.02, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1633, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210596" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:57.032Z" } }, { "rank": 8, "id": "st_20261024_393", "title": "OpenAI's Ad Pixel Tracks by Default, and European Shops Aren't Switching It Off", "url": "https://ciphercue.com/blog/openai-chatgpt-ad-pixel-european-shops", "author": null, "site": "ciphercue.com", "published_at": "2026-10-23T10:52:27.000Z", "reading_minutes": 10, "depth": "claim_checked", "format": "original_research", "summary": "A Ciphercue scan of 341 retail sites, mostly European, found OpenAI's ChatGPT ad pixel sending a measurement event before any consent on 224 of them (66%); 104 held it back. OpenAI's documentation sets consent to true by default, and the install snippet has no consent call. In the authors' harness, oaiq(\"consent\", false) before init suppressed measurement. The data covers one homepage load per site, and the decline test used one site. Anyone integrating the pixel in the EU or UK needs to add the consent gate.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "policy" ] }, "score": { "total": 0.743, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.628, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6525, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.8, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.8533, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.467, "judge_a": 0.67, "judge_b": 0.72, "mean": 0.6189, "spread": 0.253, "weight": 0.1 }, "practical_relevance": { "jev": 0.895, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.835, "spread": 0.145, "weight": 0.15 }, "substantiated": { "jev": 0.983, "judge_a": 0.85, "judge_b": 0.91, "mean": 0.9144, "spread": 0.133, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.8792, "spread": 0.198, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9267, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.48, "judge_a": 0.2, "judge_b": 0.2, "mean": 0.2933, "spread": 0.28, "weight": -0.15 } } }, "claims": [ { "text": "Of 341 shops running OpenAI's ChatGPT ad pixel, 224 (66%) send a measurement event to OpenAI before any consent, while 104 (30%) hold it back.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "66% of 341 shops running OpenAI's ChatGPT ad pixel, most of them European, send a measurement event to OpenAI before any consent (224 sites", "source_quote": null, "source_url": null }, { "text": "OpenAI's advertising business expanded to more than 60 countries in late September 2026.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "OpenAI's advertising business expanded to more than 60 countries in late September 2026", "source_quote": null, "source_url": "https://openai.com/index/new-chatgpt-ads-format-and-measurement" }, { "text": "OpenAI's documentation says the pixel initializes consent to true by default unless set to false or a stored denial is found.", "kind": "quote", "label": "no_source", "drift": null, "article_quote": "initializes consent to true by default unless you set it to false or the Pixel finds a stored denial", "source_quote": null, "source_url": null }, { "text": "The pixel's default of tracking on, left unchanged by shops, is the common cause of pre-consent firing; calling oaiq(\"consent\", false) before init suppressed all measurement events in the authors' tests.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "The common cause is the pixel's default. It ships with tracking switched on, and most of these shops never switched it off", "source_quote": null, "source_url": null }, { "text": "In September 2025 the CNIL fined the operator of shein.com 150 million euros for placing advertising cookies before any banner interaction and reading them after users clicked \"Refuse all\".", "kind": "number", "label": "matches", "drift": null, "article_quote": "In September 2025, France's data-protection regulator, the CNIL, fined the operator of the fast-fashion site shein.com 150 million euros", "source_quote": "On 1st September 2025, the CNIL imposed a fine of 150 million euros on INFINITE STYLES SERVICES CO. LIMITED, the Irish subsidiary of the SHEIN group", "source_url": "https://www.cnil.fr/en/cookies-placed-without-consent-shein-fined-150-million-euros-cnil" }, { "text": "In November 2025 the CNIL fined American Express 1.5 million euros for cookies placed before choice, despite refusal, and after consent was withdrawn.", "kind": "number", "label": "matches", "drift": null, "article_quote": "In November 2025 the same regulator fined American Express 1.5 million euros for three failures", "source_quote": "On 27th November 2025, the CNIL sanctioned AMERICAN EXPRESS CARTE FRANCE, the French subsidiary of the AMERICAN EXPRESS group, with a fine of €1.5 million for non-compliance with the rules…", "source_url": "https://www.cnil.fr/en/cookies-american-express-fined-eu15-million-cnil" }, { "text": "In September 2025 the CNIL fined Google 325 million euros over how it steered people into advertising cookies.", "kind": "number", "label": "matches", "drift": null, "article_quote": "in September 2025 the CNIL fined Google 325 million euros over how it steered people into advertising cookies", "source_quote": "On 1st September 2025, the CNIL imposed a fine of 325 million euros on Google for displaying advertisements between Gmail users' emails without their consent and for placing cookies when…", "source_url": "https://www.cnil.fr/en/cookies-and-advertisements-inserted-between-emails-google-fined-325-million-euros-cnil" }, { "text": "The UK ICO's cookie-penalty ceiling is rising from 500,000 pounds to 17.5 million pounds, or 4% of global turnover, under the Data Use and Access Act 2025.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "the ICO's cookie-penalty ceiling is rising from 500,000 pounds to 17.5 million pounds, or 4% of global turnover, under the Data Use and Access Act 2025", "source_quote": null, "source_url": "https://ico.org.uk/about-the-ico/the-data-use-and-access-act-2025" } ], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206997" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:53:57.026Z" } }, { "rank": 9, "id": "st_20261024_80", "title": "Introducing TorchTPU", "url": "https://pytorch.org/blog/introducing-torchtpu/", "author": "Ebba Simpson", "site": "PyTorch", "published_at": "2026-10-23T02:37:27.000Z", "reading_minutes": 45, "depth": "claim_checked", "format": "engineering_writeup", "summary": "Google and Meta's PyTorch team released TorchTPU, an open source PyTorch backend for TPUs that replaces the Lazy Tensor-based PyTorch/XLA. Moving a model is meant to need little more than changing the device from \"cuda\" to \"tpu\". It adds a DeferAndFuse eager mode that the authors say can run 50% or more faster than strict eager, supports about 450 ATen ops, and works with vLLM, SGLang and TorchTitan. The performance, model-coverage and benchmark figures come from the authors and were not independently verified.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "programming", "ai", "hardware" ] }, "score": { "total": 0.7419, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.71, "judge_a": 0.72, "judge_b": 0.85, "mean": 0.76, "spread": 0.14, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.84, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.6, "judge_a": 0.74, "judge_b": 0.72, "mean": 0.6867, "spread": 0.14, "weight": 0.1 }, "practical_relevance": { "jev": 0.785, "judge_a": 0.78, "judge_b": 0.9, "mean": 0.8217, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.63, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7033, "spread": 0.15, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.92, "judge_b": 1, "mean": 0.97, "spread": 0.08, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 1, "mean": 0.9367, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.1, "judge_a": 0.12, "judge_b": 0.18, "mean": 0.1333, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.48, "judge_a": 0.22, "judge_b": 0.28, "mean": 0.3267, "spread": 0.26, "weight": -0.15 } } }, "claims": [ { "text": "TorchTPU started as an effort at Google in mid 2025, in collaboration with the PyTorch Team at Meta, to run PyTorch natively on TPUs.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "a new effort started at Google in mid 2025, in collaboration with the PyTorch Team at Meta, to run PyTorch natively on TPUs", "source_quote": null, "source_url": null }, { "text": "TorchTPU provides out-of-box execution on the 5,000 most downloaded Hugging Face models, validated with Transformers v5, Diffusers and Kernels libraries.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "with out-of-box execution on the 5,000 most downloaded Hugging Face models", "source_quote": null, "source_url": "https://huggingface.co/blog/hf-torchtpu" }, { "text": "The DeferAndFuse eager mode can be 50% or more faster than Strict Eager while reducing peak HBM memory.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "can be 50% or more faster than Strict Eager while reducing peak HBM memory", "source_quote": null, "source_url": null }, { "text": "In the Hugging Face model coverage test, 98% (4,821 of 4,917) of Tier 1 models passed without crashing, and 97% (638 of 657) of Tier 2 architecture representatives achieved numerical parity with GPU.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "98% – 4,821 out of 4,917", "source_quote": null, "source_url": null }, { "text": "With TorchTitan, Qwen3 MoE 235B on 64 TPU v7 chips achieved 1828 TPS with PyTorch/TorchTitan versus 1937 TPS with JAX/MaxText.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "1828 TPS for PyTorch/TorchTitan vs 1937 TPS for JAX/MaxText", "source_quote": null, "source_url": null }, { "text": "TPU v7 with the vLLM-TorchTPU stack reached the lowest cost per million tokens overall on InferenceX 8K/1K at 2.01 cents, down from 3.6 cents at the September debut.", "kind": "benchmark", "label": "couldnt_check", "drift": null, "article_quote": "lowest cost per million tokens overall at 2.01 cents (down from 3.6 cents in our September debut", "source_quote": null, "source_url": "https://newsletter.semianalysis.com/p/tpu-inferencex-full-steam" }, { "text": "SemiAnalysis's independent benchmark found TPU v7 with the vLLM-TorchTPU stack delivered industry-competitive performance-per-dollar.", "kind": "causal", "label": "couldnt_check", "drift": null, "article_quote": "Google’s TPU v7 paired with the vLLM-TorchTPU stack, delivered industry-competitive performance-per-dollar", "source_quote": null, "source_url": "https://newsletter.semianalysis.com/p/tpu-inferencex-full-steam" }, { "text": "On a 32-chip TPU v7x cluster with DLRM v2, the TorchRec TPU backend matched Google's JAX baseline with an identical 18.3ms step latency.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "achieved 100% performance parity with Google’s highly optimized JAX baseline, reaching an identical 18.3ms step latency", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203860" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:53:57.152Z" } }, { "rank": 10, "id": "st_20261024_1219", "title": "Watching the Watchers: Analysis of U.S. Smart TV Automatic Content Recognition", "url": "https://watchingthewatchers.github.io/watchingthewatchers/index.html", "author": null, "site": "watchingthewatchers.github.io", "published_at": "2026-10-23T19:17:13.000Z", "reading_minutes": 11, "depth": "three_judges", "format": "original_research", "summary": "Researchers tested 12 smart TVs covering eight operating systems, running 5,943 automated experiments with ACR on and after opting out. Antenna viewing was tracked on every analyzable OS, while Netflix and YouTube were not. Opting out fully worked on four systems; on Samsung and Vizio, traffic consistent with ACR continued. VIDAA OS sent ACR data unencrypted. Most ACR traffic is encrypted, so findings rely on traffic behavior, and results are limited to these models and firmware.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "security", "hardware", "policy" ] }, "score": { "total": 0.7407, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.728, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.6992, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.8133, "spread": 0.25, "weight": 0.2 }, "audience_breadth": { "jev": 0.65, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.59, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.885, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.7383, "spread": 0.335, "weight": 0.15 }, "substantiated": { "jev": 0.983, "judge_a": 0.85, "judge_b": 0.91, "mean": 0.9144, "spread": 0.133, "weight": 0.1 }, "technical_depth": { "jev": 0.933, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.8575, "spread": 0.132, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9233, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.2, "judge_b": 0.16, "mean": 0.2067, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50214005" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:57.160Z" } }, { "rank": 11, "id": "st_20261024_620", "title": "Building an Isolated Preview for AI-Generated HTML", "url": "https://blog.master.dev/building-an-isolated-preview-for-ai-generated-html/", "author": "Keshav Malik October 22, 2026", "site": "blog.master.dev", "published_at": "2026-10-23T13:19:08.000Z", "reading_minutes": 10, "depth": "three_judges", "format": "engineering_writeup", "summary": "A Master.dev tutorial shows how to build an isolated preview for model-generated HTML, using demos in which an unsandboxed srcdoc iframe reads the host page's token. A sandbox=\"allow-scripts\" frame cannot read the host or navigate the tab, but it can still send fetch and image requests out. A default-src 'none' CSP blocks those requests. The author advises against combining allow-scripts with allow-same-origin, hosting previews on a throwaway origin, and validating postMessage by event.source, since sandboxed senders report \"null\". The demos are browser-based and the claims were not independently checked.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "ai", "security" ] }, "score": { "total": 0.7403, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.638, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5858, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.8, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8333, "spread": 0.1, "weight": 0.2 }, "audience_breadth": { "jev": 0.673, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6644, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.983, "judge_a": 0.85, "judge_b": 0.93, "mean": 0.9208, "spread": 0.133, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8278, "spread": 0.263, "weight": 0.1 }, "technical_depth": { "jev": 1, "judge_a": 0.85, "judge_b": 0.88, "mean": 0.91, "spread": 0.15, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.5, "judge_b": 0.98, "mean": 0.8033, "spread": 0.48, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.18, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208551" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:57.276Z" } }, { "rank": 12, "id": "st_20261024_37", "title": "Red-Teaming OpenAI's Internal Agent Monitoring Systems", "url": "https://blog.redwoodresearch.org/p/red-teaming-openais-internal-agent", "author": "Buck Shlegeris", "site": "blog.redwoodresearch.org", "published_at": "2026-10-23T01:41:30.000Z", "reading_minutes": 6, "depth": "claim_checked", "format": "engineering_writeup", "summary": "METR and Redwood Research staff spent three weeks in July and August with deep internal access to OpenAI, testing its asynchronous misalignment monitor as it stood on July 21. They estimate 75–94% of agentic Responses API traffic was eligible for monitoring, and roughly 10% of eligible Codex traffic went unmonitored through compaction gaps. They also found client-controllable headers that disable monitoring and serialization mismatches that could make the monitor fail. OpenAI says the issues are patched; the authors have not verified this, and the public write-up is a redacted summary.", "also_covered_by": [ { "site": "The Register", "url": "https://www.theregister.com/ai-and-ml/2026/10/23/red-team-report-found-openai-utterly-failed-to-monitor-its-agents-during-hugging-face-attack/5305200", "title": "Red-team report found OpenAI utterly failed to monitor its agents during Hugging Face attack" } ], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.7353, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.68, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.7133, "spread": 0.26, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.9, "judge_b": 0.97, "mean": 0.8667, "spread": 0.24, "weight": 0.2 }, "audience_breadth": { "jev": 0.34, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.5533, "spread": 0.38, "weight": 0.1 }, "practical_relevance": { "jev": 0.698, "judge_a": 0.7, "judge_b": 0.83, "mean": 0.7425, "spread": 0.132, "weight": 0.15 }, "substantiated": { "jev": 0.78, "judge_a": 0.75, "judge_b": 0.87, "mean": 0.8, "spread": 0.12, "weight": 0.1 }, "technical_depth": { "jev": 0.748, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.7625, "spread": 0.14, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.95, "judge_b": 0.99, "mean": 0.9567, "spread": 0.06, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0467, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.15, "judge_b": 0.12, "mean": 0.1867, "spread": 0.17, "weight": -0.15 } } }, "claims": [ { "text": "A METR employee and a Redwood Research employee conducted a three-week pilot high-access exercise, from July 21st to August 11th, to stress-test OpenAI's internal misalignment monitoring system.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "For three weeks, from July 21st to August 11th, a METR employee and a Redwood Research employee conducted a pilot high-access exercise", "source_quote": null, "source_url": null }, { "text": "The exercise produced a 61-page internal report that was shared with OpenAI researchers and security staff.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "The output of this exercise was a 61-page internal report that was shared with OpenAI researchers and security staff", "source_quote": null, "source_url": null }, { "text": "METR and Redwood estimated that roughly 75–94% of agentic internal Responses API traffic from models at least as strong as GPT-5.5 was eligible for monitoring at the time of the exercise.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "roughly 75–94% of agentic internal Responses API (which powers Codex, among other things)4 traffic from models at least as strong as GPT-5.5 was eligible for monitoring", "source_quote": null, "source_url": null }, { "text": "Of the Responses API traffic used for internal Codex agents, 99.2% was eligible for monitoring.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "out of the subset of Responses API traffic used for internal Codex agents, 99.2% was eligible for monitoring", "source_quote": null, "source_url": null }, { "text": "Not monitoring compaction events likely led to on the order of 10% of eligible internal Codex traffic not being monitored between May and July 2026.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "not monitoring compaction events likely led to on the order of 10% of eligible internal Codex traffic not being monitored between May and July 2026", "source_quote": null, "source_url": null }, { "text": "Information given to the researchers suggested the monitor errored on less than 0.2% of requests, with zero load-sheds in one 24-hour period during the exercise.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "the monitor errored on less than 0.2% of requests, and that in one 24-hour period during the exercise there were zero load-sheds", "source_quote": null, "source_url": null }, { "text": "The researchers found two independent client-controllable headers that disable the monitoring system for new agents.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "We found two independent client-controllable headers that disable the monitoring system for new agents", "source_quote": null, "source_url": null }, { "text": "OpenAI has stated that all vulnerabilities identified that affect its production safeguards stack have been patched.", "kind": "quote", "label": "couldnt_check", "drift": null, "article_quote": "OpenAI has stated that all of the vulnerabilities we identified that impact their production safeguards stack have been patched", "source_quote": null, "source_url": "https://openai.com/index/hugging-face-incident-and-the-road-ahead" } ], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203538" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:57.290Z" } }, { "rank": 13, "id": "st_20261024_507", "title": "openssl/openssl openssl-4.1.0", "url": "https://github.com/openssl/openssl/releases/tag/openssl-4.1.0", "author": "openssl-machine", "site": "GitHub", "published_at": "2026-10-23T12:23:31.000Z", "first_published_at": "2026-10-23T12:23:31.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "release", "summary": "OpenSSL 4.1.0 adds DTLS 1.3 (RFC 9147), RFC 8701 GREASE, DTLS in the SSL listener API, IKEv2 KDF and initial Elbrus2000 support. It also adds AVX-512 and VAES speedups for ML-DSA, SHAKE and AES-CBC decryption on x86_64, plus ML-DSA and ML-KEM optimizations on ppc64le. Itanium and Windows CE targets and the no-ecdsa/no-ecdh options are removed. Known issues include DTLS 1.3 KeyUpdate dropping pending retransmissions and DTLS listener error-reporting bugs, so teams should test before adopting DTLS 1.3.", "also_covered_by": [ { "site": "Phoronix", "url": "https://www.phoronix.com/news/OpenSSL-4.1-Released", "title": "OpenSSL 4.1 Released With More AVX-512 Optimizations" } ], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "security" ] }, "score": { "total": 0.7268, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.738, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.7025, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.53, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.7033, "spread": 0.35, "weight": 0.2 }, "audience_breadth": { "jev": 0.533, "judge_a": 0.67, "judge_b": 0.78, "mean": 0.6611, "spread": 0.247, "weight": 0.1 }, "practical_relevance": { "jev": 0.84, "judge_a": 0.7, "judge_b": 0.8, "mean": 0.78, "spread": 0.14, "weight": 0.15 }, "substantiated": { "jev": 0.677, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.8089, "spread": 0.273, "weight": 0.1 }, "technical_depth": { "jev": 0.725, "judge_a": 0.7, "judge_b": 0.7, "mean": 0.7083, "spread": 0.025, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.9, "judge_b": 1, "mean": 0.9467, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0467, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.1, "judge_a": 0.1, "judge_b": 0.04, "mean": 0.08, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 184, "front_page": true, "url": "https://news.ycombinator.com/item?id=50208099" }, "lobsters": null, "line": "Elsewhere: HN 184 pts, front page", "checked_at": "2026-10-24T08:53:57.282Z" } }, { "rank": 14, "id": "st_20261024_228", "title": "GLM-5.3 Optimizations, Part 2: PCP, DCP, and Hybrid HiSparse on 16 H200s", "url": "https://vllm.ai/blog/2026-10-23-glm53-part2-pcp-dcp-hisparse", "author": "vLLM Team", "site": "vLLM Blog", "published_at": "2026-10-23T00:00:00.000Z", "first_published_at": "2026-10-23T00:00:00.000Z", "reading_minutes": 9, "depth": "claim_checked", "format": "engineering_writeup", "summary": "vLLM's second GLM-5.3 write-up serves long agentic contexts on 16 H200s split into prefill and decode nodes. Prefill combines token-wise context parallelism (PCP8) with KV sharding (DCP8), and decode uses Hybrid HiSparse host-memory offload. On AgentX, the authors report 1.7–3.2× the per-user speed of data-parallel prefill at equal throughput, median time to first token of 0.8–0.9 s, and the only tested setup serving 1M-token contexts. Results are self-reported, from one benchmark with fixed speculative-decoding acceptance, and DCP adds 6–15% per-step overhead.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "programming", "ai", "devops" ] }, "score": { "total": 0.7264, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.64, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.64, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.76, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8233, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.357, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.4456, "spread": 0.223, "weight": 0.1 }, "practical_relevance": { "jev": 0.798, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7858, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.957, "judge_a": 0.85, "judge_b": 0.92, "mean": 0.9089, "spread": 0.107, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.95, "judge_b": 0.98, "mean": 0.9758, "spread": 0.048, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9233, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.2, "judge_b": 0.04, "mean": 0.0967, "spread": 0.16, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.18, "spread": 0.18, "weight": -0.15 } } }, "claims": [ { "text": "On AgentX, PCP8 × DCP8 prefill with Hybrid HiSparse decode gives 1.7–3.2× the per-user speed of a data-parallel deployment at the same throughput.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "PCP8 × DCP8 prefill with Hybrid HiSparse decode gives 1.7–3.2× the per-user speed of a data-parallel deployment at the same throughput", "source_quote": null, "source_url": null }, { "text": "PCP8 × DCP8 with Hybrid HiSparse is the only configuration the authors tested that serves 1M-token contexts.", "kind": "superlative", "label": "no_source", "drift": null, "article_quote": "it is the only configuration we tested that serves 1M-token contexts", "source_quote": null, "source_url": null }, { "text": "PCP8 × DCP8 prefill cuts median TTFT from 1.5–5.2 s (DP8) to 0.8–0.9 s.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "cuts median TTFT from 1.5–5.2 s to 0.8–0.9 s", "source_quote": null, "source_url": null }, { "text": "At c64, HiSparse decode nearly doubles throughput to 8,340 logical tok/s/GPU at 29.6 tok/s/user, while every other configuration drops to 1–2 tok/s/user.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "At c64 it nearly doubles throughput, to 8,340 logical tok/s/GPU at 29.6 tok/s/user, while every other configuration drops to 1–2 tok/s/user", "source_quote": null, "source_url": null }, { "text": "A 1M-token GLM-5.3 request needs 53.4 GiB of FP8 KV cache, while each GPU has room for only about 15 GiB; DCP8 reduces this to 6.7 GiB per GPU.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "A 1M-token GLM-5.3 request needs 53.4 GiB of FP8 KV,3 but each GPU has room for only about 15 GiB", "source_quote": null, "source_url": null }, { "text": "DCP8 adds about 6% per step overhead up to 64K tokens of context, rising to 15% at 248K.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "DCP8 adds about 6% per step up to 64K tokens of context, rising to 15% at 248K", "source_quote": null, "source_url": null }, { "text": "PCP raises sparse MLA's arithmetic intensity from about 26 to about 191 FLOPs per byte moved.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "raising sparse MLA from about 26 to about 191 FLOPs per byte moved", "source_quote": null, "source_url": null }, { "text": "The indexer overtakes sparse MLA under TP8 at about 32K tokens, while under PCP8 the crossover moves out to about 256K.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "the indexer overtakes sparse MLA under TP8 once the prompt reaches about 32K tokens; under PCP8, where no GPU replicates it, the crossover moves out to about 256K", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:57.539Z" } }, { "rank": 15, "id": "st_20261024_452", "title": "Azure CLI still trusts compromised eMudhra CA; fix targeted for November 16", "url": "https://github.com/Azure/azure-cli/pull/34250", "author": "Copilot", "site": "GitHub", "published_at": "2026-10-23T11:47:27.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "news_report", "summary": "Azure CLI bundles certifi 2024.7.4, which still trusts the eMudhra CA certificates whose private keys were publicly compromised (GHSA-6q48-6c7w-9gw2, CVSS 9.1). The PR author says anyone on the same network can forge certificates for Azure endpoints and, through the extension index, serve malicious extensions. A three-line pin bump is approved but milestoned for November 16, weeks after the October 20 disclosure. Until then, MSI, deb, rpm, Docker and Homebrew users can set REQUESTS_CA_BUNDLE to a current certifi bundle.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "established_project", "categories": [ "programming", "security", "devops" ] }, "score": { "total": 0.7254, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.69, "judge_a": 0.55, "judge_b": 0.86, "mean": 0.7, "spread": 0.31, "weight": 0.25 }, "reader_interest": { "jev": 0.83, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.86, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.437, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.5856, "spread": 0.283, "weight": 0.1 }, "practical_relevance": { "jev": 0.848, "judge_a": 0.85, "judge_b": 0.91, "mean": 0.8692, "spread": 0.063, "weight": 0.15 }, "substantiated": { "jev": 0.773, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.7711, "spread": 0.14, "weight": 0.1 }, "technical_depth": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.79, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8333, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0267, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.44, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.3067, "spread": 0.26, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 6, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207472" }, "lobsters": null, "line": "Elsewhere: HN 6 pts, not on front page", "checked_at": "2026-10-24T08:53:57.398Z" } }, { "rank": 16, "id": "st_20261024_1562", "title": "Harm Is Behavioural, Moderation Is Lexical: How Language and Culture Decide Whom Automated Safety Systems Protect and Whom They Silence", "url": "https://arxiv.org/abs/2610.31047", "author": "Gabriel Lars Sabadin, Dr Nicola Harding, Sarah Kingston", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T10:24:17.000Z", "reading_minutes": 41, "depth": "claim_checked", "format": "original_research", "summary": "A new arXiv paper argues that moderation systems judge harm by vocabulary rather than by what a message does, which both over-flags dialects and under-protects low-resource languages. In the authors' test of five open systems across eleven languages, 51 of 55 system–language pairs fell short of native-speaker accuracy, and Arabic and romanised Hindi stayed near chance even as capability rose. The behaviour-based alternative is stated as hypotheses, not tested. Several cited figures could not be independently verified, and the authors' own results are not yet sourced.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "policy" ] }, "score": { "total": 0.7239, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.755, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.7083, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.75, "judge_a": 0.72, "judge_b": 0.88, "mean": 0.7833, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.64, "judge_a": 0.62, "judge_b": 0.78, "mean": 0.68, "spread": 0.16, "weight": 0.1 }, "practical_relevance": { "jev": 0.865, "judge_a": 0.66, "judge_b": 0.82, "mean": 0.7817, "spread": 0.205, "weight": 0.15 }, "substantiated": { "jev": 0.973, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8778, "spread": 0.173, "weight": 0.1 }, "technical_depth": { "jev": 0.983, "judge_a": 0.78, "judge_b": 0.9, "mean": 0.8875, "spread": 0.203, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8333, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.1, "judge_b": 0.01, "mean": 0.05, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.47, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.3167, "spread": 0.29, "weight": -0.15 } } }, "claims": [ { "text": "Sap et al. (2019) found that tweets in African American English, and tweets by self-identified African Americans, were up to two times more likely to be labelled offensive by models trained on widely used hate-speech datasets.", "kind": "number", "label": "matches", "drift": null, "article_quote": "tweets in African American English, and tweets by self-identified African Americans, were up to two times more likely to be labelled offensive", "source_quote": "models trained on these corpora acquire and propagate these biases, such that AAE tweets and tweets by self-identified African Americans are up to two times more likely to be labelled…", "source_url": "https://doi.org/10.18653/v1/P19-1163" }, { "text": "Dias Oliva, Antonialli and Gomes (2021) found that a significant number of prominent drag queens' accounts scored as more toxic than those of white nationalists on a widely used toxicity-scoring service.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "found that a significant number of their accounts scored as more toxic than those of white nationalists", "source_quote": null, "source_url": "https://doi.org/10.1007/s12119-020-09790-w" }, { "text": "Yong, Menghini and Bach (2023) showed that translating unsafe English prompts into low-resource languages raised the rate at which a widely deployed model's safeguards were bypassed from under 1% to 79%.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "raised the rate at which a widely deployed model’s safeguards were bypassed from under 1% to 79%", "source_quote": null, "source_url": null }, { "text": "Deng et al. (2024) found that low-resource languages were about three times as likely to elicit harmful content as high-resource ones, even without adversarial intent.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "low-resource languages were about three times as likely to elicit harmful content as high-resource ones, even without adversarial intent", "source_quote": null, "source_url": null }, { "text": "Röttger et al. report that all four models tested classified most non-hateful reclaimed-slur cases as hateful, with accuracies between roughly 18% and 40%, the commercial services lowest.", "kind": "benchmark", "label": "couldnt_check", "drift": null, "article_quote": "with accuracies between roughly 18% and 40% and the commercial services lowest", "source_quote": null, "source_url": "https://doi.org/10.18653/v1/2021.acl-long.4" }, { "text": "In the authors' worked example, 51 of 55 system–language combinations fall reliably short of native-speaker annotators' balanced accuracy at separating implicit hate from harmless contrast cases, by 6 to 55 points.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "Of the 55 combinations, 51 fall reliably short of that standard, by 6 to 55 points", "source_quote": null, "source_url": null }, { "text": "Averaged over nine languages, AUC rises from 0.65 for the small Qwen3Guard to 0.81 for both the 8B model and PolyGuard, while Arabic and romanised Hindi barely move, so the spread of discrimination across languages widens with capability.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "The spread of discrimination across languages therefore widens with capability, from 0.20 for the small guard model to 0.29 for the 8B model and 0.33 for PolyGuard", "source_quote": null, "source_url": null }, { "text": "In a stratified percentile bootstrap simulation, the interval never covered a true gap of zero and covered a true gap of 10 points in only 64–76% of replicates against a nominal 95%.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "covered a true gap of 10 points in only 64–76% of replicates, against a nominal 95%", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:57.629Z" } }, { "rank": 17, "id": "st_20261024_1314", "title": "Investigating the PingCAP Security Incident", "url": "https://nebusec.ai/research/pingcap-security-incident/", "author": null, "site": "NebuSec", "published_at": "2026-10-23T21:02:27.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "engineering_writeup", "summary": "NebuSec's investigation of the PingCAP breach (September 9 to October 7, 2026) found an attacker entered through a compromised GitHub account, then used credentials in repositories and Terraform state to reach Google Cloud, Jira and internal chat. Exported data included 6,458 BigQuery tables, source code, billing records and support tickets containing customer-shared AWS keys. NebuSec found no evidence that live TiDB databases were accessed. The case shows how plaintext secrets and support-ticket credentials widen a breach.", "also_covered_by": [ { "site": "TiDB", "url": "https://www.pingcap.com/trust-hub/security/ceo-note-october-2026/", "title": "A Note from Our CEO on the PingCAP Security Incident" } ], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "security", "devops", "it", "data" ] }, "score": { "total": 0.7195, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.623, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6508, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.82, "judge_b": 0.95, "mean": 0.8533, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.337, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.5522, "spread": 0.383, "weight": 0.1 }, "practical_relevance": { "jev": 0.91, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.84, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.817, "judge_a": 0.72, "judge_b": 0.9, "mean": 0.8122, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.865, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.815, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9367, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.35, "judge_b": 0.08, "mean": 0.16, "spread": 0.3, "weight": -0.15 }, "overclaim": { "jev": 0.13, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.1833, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50215283" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:57.413Z" } }, { "rank": 18, "id": "st_20261024_1623", "title": "What the Specification Did Not Say: Specification-Oracle Gaps in Test-Graded Coding-Agent Benchmarks", "url": "https://arxiv.org/abs/2610.30635", "author": "Fei Ding", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T02:56:44.000Z", "reading_minutes": 37, "depth": "claim_checked", "format": "original_research", "summary": "An audit of SWE-Milestone, where coding agents are graded with upstream project tests, finds that 25 of 98 milestones include a test requiring a name or path the specification never states. Build units with such a dependency fail in 80.6% of runs, against 48.4% for comparable units. Removing the affected tests leaves the ranking within chance variation but shifts go-zero scores. Audit precision is 89.2% on held-out cases, and the detailed findings cover one benchmark of seven repositories.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.7187, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.678, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6358, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.75, "judge_a": 0.7, "judge_b": 0.94, "mean": 0.7967, "spread": 0.24, "weight": 0.2 }, "audience_breadth": { "jev": 0.57, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5833, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.685, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7017, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8978, "spread": 0.193, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.9067, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8933, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1867, "spread": 0.12, "weight": -0.15 } } }, "claims": [ { "text": "On SWE-Milestone, 25 of the 98 graded milestones contain a test that depends on an unstated, non-inferable name or path.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "25 of the 98 graded milestones contain a required test with a hard hidden dependency", "source_quote": null, "source_url": null }, { "text": "Build units with a hard hidden dependency fail in 80.6% of (unit, run) pairs, against 48.4% for comparable units without one.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "units blocked by a hard hidden dependency fail in 80.6% of (unit, run) pairs", "source_quote": null, "source_url": null }, { "text": "Renaming one hidden identifier in the reference solution stops the unit's tests from compiling in 126 of 128 Go cases.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "stops the unit’s tests from compiling in 126 of 128 Go cases", "source_quote": null, "source_url": null }, { "text": "The static audit has precision 89.2% on flagged units and recall 92.9% on the second held-out set.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "the audit has precision 58/65 (89.2%; 95% Wilson interval 79.4–94.7%); on the second set its recall is 13/14 (92.9%", "source_quote": null, "source_url": null }, { "text": "Removing the affected tests leaves the leaderboard ranking essentially unchanged (Kendall's tau-b 0.984 by score and 0.942 by resolve rate for the 28 leaderboard runs in the full variant).", "kind": "generalization", "label": "no_source", "drift": null, "article_quote": "Kendall’s $\\tau$ -b with the published ranking is 0.984 by score and 0.942 by resolve rate", "source_quote": null, "source_url": null }, { "text": "SWE-bench Verified removed tasks whose tests checked details absent from the issue.", "kind": "causal", "label": "couldnt_check", "drift": null, "article_quote": "SWE-bench Verified (OpenAI, 2024) removed tasks whose tests checked details absent from the issue", "source_quote": null, "source_url": "https://openai.com/index/introducing-swe-bench-verified" }, { "text": "Runs reproduce 45% of the hard hidden dependencies with evidence and 79% of the inferable ones, with no sign that reproduction depends on training cutoff.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "Runs reproduce 45% of the hard dependencies for which their evaluation gives evidence and 79% of the inferable ones", "source_quote": null, "source_url": null }, { "text": "Prevalence of gaps follows how the task text is made: a third or more of tasks are affected where specifications are release notes or LLM summaries, almost none where tasks are screened or written for their tests.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "Prevalence follows construction", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:57.753Z" } }, { "rank": 19, "id": "st_20261024_660", "title": "Trust Stores Are Dependencies Too", "url": "https://nesbitt.io/2026/10/21/trust-stores-are-dependencies-too.html", "author": "Andrew Nesbitt", "site": "Andrew Nesbitt", "published_at": "2026-10-23T14:52:08.000Z", "first_published_at": "2026-10-21T00:00:00.000Z", "reading_minutes": 7, "depth": "claim_checked", "format": "engineering_writeup", "summary": "After eMudhra mis-generated eight intermediate CA certificates, whose private keys can be recovered, browsers and OS vendors blocked them, but clients without revocation checks need package updates. Andrew Nesbitt scanned 6,877 critical packages and found 29 shipping CA bundles; 26 hold Mozilla-removed roots still within validity. His counts are unverified, and the scanner misses compiled-in roots. The Apple block date is disputed (19th versus 22nd). Lockfiles, SBOMs and registry metadata don't record bundled trust stores, so finding them means parsing archive contents.", "also_covered_by": [ { "site": "Andrew Nesbitt", "url": "https://nesbitt.io/2026/10/23/one-small-bug.html", "title": "One Small Bug" } ], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "security", "devops", "it" ] }, "score": { "total": 0.7171, "contested": false, "flag_penalty": 0.1, "breakdown": { "significance": { "jev": 0.73, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7167, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.82, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.8733, "spread": 0.13, "weight": 0.2 }, "audience_breadth": { "jev": 0.937, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8056, "spread": 0.237, "weight": 0.1 }, "practical_relevance": { "jev": 0.858, "judge_a": 0.85, "judge_b": 0.91, "mean": 0.8725, "spread": 0.06, "weight": 0.15 }, "substantiated": { "jev": 0.933, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.9078, "spread": 0.09, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.9, "judge_b": 0.93, "mean": 0.9408, "spread": 0.093, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.91, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0233, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.1367, "spread": 0.1, "weight": -0.15 } } }, "claims": [ { "text": "eMudhra generated eight of its intermediate CA certificates incorrectly, so their private keys can be recovered from the certificates themselves.", "kind": "causal", "label": "matches", "drift": null, "article_quote": "eMudhra generated eight of its intermediate CA certificates incorrectly, so their private keys can be recovered from the certificates themselves", "source_quote": "The private keys for the intermediate certificates were incorrectly generated by the CA. This means that anyone can recover them from the certificates", "source_url": "https://emudhra.we-have-a-problem.com/" }, { "text": "Firefox and Chrome blocked the eMudhra intermediates on 17 October, Apple on the 19th, and Microsoft on the 20th.", "kind": "date", "label": "differs", "drift": null, "article_quote": "Firefox (OneCRL and CRLite) and Chrome (CRLSets) blocked the intermediates on 17 October, Apple on the 19th, and Microsoft on the 20th", "source_quote": "The certificates have been blocked in Firefox and Chrome since the 17th of October, on Microsoft platforms since the 20th and on Apple products since the 22nd", "source_url": "https://emudhra.we-have-a-problem.com/" }, { "text": "Of the 6,877 critical packages scanned, 116 contained certificates (about 1.7%), and 28 ship something that works as a CA bundle.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "116 of those packages contained certificates, about 1.7%, mostly test fixtures and package-signing certs. 28 of them ship something that works as a CA bundle", "source_quote": null, "source_url": "https://gist.github.com/andrew/4ee27639efb1eb539f36e791942e6617" }, { "text": "26 of the 29 bundle packages contain at least one root that Mozilla removed from its programme and that is still within its validity period; matching against CCADB gives 95 such roots.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "26 of the 29 bundle packages contain at least one root that Mozilla has removed from its programme and that is still inside its validity period", "source_quote": null, "source_url": "https://ccadb.my.salesforce-sites.com/ccadb/AllCertificateRecordsCSVFormatV5" }, { "text": "Mozilla removed emSign ECC Root CA - G3 in response to the incident, as its 20 October notice records.", "kind": "causal", "label": "matches", "drift": null, "article_quote": "Mozilla removed emSign ECC Root CA - G3 in response to the incident, as its 20 October notice records", "source_quote": "The G3 root removal is in response to a recent security incident which impacted some intermediate certificates", "source_url": "https://groups.google.com/a/mozilla.org/g/dev-tech-crypto/c/vSxZxnOMMvs" }, { "text": "certifi carried emSign ECC Root CA - G3 through 2026.7.22 and dropped it in 2026.10.18, a day before the 2026.10.19 release.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "certifi carried it through 2026.7.22 and dropped it in 2026.10.18, a day before the 2026.10.19 release in the fix table", "source_quote": null, "source_url": null }, { "text": "Algoliasearch, aws-sdk-v1 and Guzzle 3 still contain StartCom roots that Mozilla removed in Firefox 58 in 2017.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "still contain StartCom roots that Mozilla removed in Firefox 58 in 2017, and those roots stay valid until 2036 and 2039", "source_quote": null, "source_url": "https://blog.mozilla.org/security/2017/08/30/removing-disabled-wosign-startcom-certificates-firefox-58" }, { "text": "Excon replaced its bundle in version 1.8.0, released on 21 October.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Excon replaced its bundle in 1.8.0, released on 21 October", "source_quote": null, "source_url": "https://github.com/excon/excon/pull/925" } ], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210069" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:57.544Z" } }, { "rank": 20, "id": "st_20261024_894", "title": "LLM Multi-Agent Systems Are Vulnerable to Herding and Runaway Cascades", "url": "https://arxiv.org/abs/2610.30769", "author": "Farnoosh Hashemi, Shiyu Ji, Joice Chen, Michael W. Macy", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T05:42:55.000Z", "reading_minutes": 62, "depth": "three_judges", "format": "original_research", "summary": "Researchers at Cornell and Stanford reran the Music Lab experiment with 13 LLM populations (168,480 agents) recommending 48 unfamiliar songs, with and without visible cumulative counts. Showing counts pushed recommendations away from independent judgment in every model, by about nine times the human benchmark under weak influence and four times under strong. Non-reasoning models diverged more and were less predictable than reasoning models, and higher capability scores correlated with less herding. The human comparison is qualitative, since the tasks differed.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.7138, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.718, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6825, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8067, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.807, "judge_a": 0.65, "judge_b": 0.72, "mean": 0.7256, "spread": 0.157, "weight": 0.1 }, "practical_relevance": { "jev": 0.665, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6617, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8767, "spread": 0.17, "weight": 0.1 }, "technical_depth": { "jev": 0.943, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.8142, "spread": 0.343, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.44, "judge_a": 0.25, "judge_b": 0.2, "mean": 0.2967, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211188" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:57.662Z" } }, { "rank": 21, "id": "st_20261024_690", "title": "Binary read replicas in PostgreSQL, Part 2", "url": "https://thebuild.com/blog/binary-read-replicas-in-postgresql-part-2/", "author": null, "site": "thebuild.com", "published_at": "2026-10-23T14:21:34.000Z", "reading_minutes": 19, "depth": "claim_checked", "format": "tutorial", "summary": "This Part 2 cookbook on PostgreSQL binary read replicas covers how to route read traffic to replicas, and why a replica should be tuned for a single job. It recommends application-side routing over middleware, uses libpq's target_session_attrs to pick standby servers, and advises health checks that include replication lag.", "also_covered_by": [], "labels": { "content_type": "tutorial", "project_stage": "not_a_project", "categories": [ "devops", "data" ] }, "score": { "total": 0.7137, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.555, "judge_a": 0.35, "judge_b": 0.78, "mean": 0.5617, "spread": 0.43, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.72, "judge_b": 0.93, "mean": 0.8133, "spread": 0.21, "weight": 0.2 }, "audience_breadth": { "jev": 0.383, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.4544, "spread": 0.197, "weight": 0.1 }, "practical_relevance": { "jev": 0.925, "judge_a": 0.85, "judge_b": 0.91, "mean": 0.895, "spread": 0.075, "weight": 0.15 }, "substantiated": { "jev": 0.583, "judge_a": 0.8, "judge_b": 0.83, "mean": 0.7378, "spread": 0.247, "weight": 0.1 }, "technical_depth": { "jev": 1, "judge_a": 0.95, "judge_b": 0.98, "mean": 0.9767, "spread": 0.05, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.5, "judge_b": 0.99, "mean": 0.81, "spread": 0.49, "weight": 0.1 }, "promotional": { "jev": 0.02, "judge_a": 0.02, "judge_b": 0.01, "mean": 0.0167, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.11, "judge_a": 0.15, "judge_b": 0.12, "mean": 0.1267, "spread": 0.04, "weight": -0.15 } } }, "claims": [ { "text": "In libpq, target_session_attrs=standby arrived in PostgreSQL 14 and load_balance_hosts=random in PostgreSQL 16.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "The first arrived in PostgreSQL 14 and the second in 16, but both are features of the client library", "source_quote": null, "source_url": null }, { "text": "Patroni's REST API endpoint GET /replica returns 200 only from a running replica, and GET /replica?lag=16MB returns 503 from a replica further behind than the threshold, measured in bytes of WAL.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "GET /replica returns 200 only from a running replica", "source_quote": null, "source_url": null }, { "text": "pgBackRest has a backup-standby option that copies most of the files from a replica, while the backup is still started and stopped on the primary.", "kind": "number", "label": "matches", "drift": null, "article_quote": "pgBackRest has a backup-standby option that copies most of the files from a replica (the backup is still started and stopped on the primary, so both have to be configured", "source_quote": "Backups can be made from a standby to reduce load on the primary. Only the few files that are not replicated are copied from the primary", "source_url": "https://github.com/pgbackrest/pgbackrest" }, { "text": "Since PostgreSQL 16, a logical replication slot can be created on a replica and logical decoding can run there.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "Since PostgreSQL 16, a logical replication slot can be created on a replica, and logical decoding can run there", "source_quote": null, "source_url": null }, { "text": "transaction_timeout is available in PostgreSQL 17 and later.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "transaction_timeout (PostgreSQL 17 and later", "source_quote": null, "source_url": null }, { "text": "A replica running a different operating system release from its primary can sort strings differently, making its block-for-block copied text indexes wrong on the replica so queries miss rows.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "A replica running a different release from its primary can disagree with the primary about how strings sort", "source_quote": null, "source_url": null }, { "text": "Through a chain of replicas that all have hot_standby_feedback on, a long query at the bottom holds back vacuum on the primary however many hops away it is.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "a long query at the bottom holds back vacuum on the primary however many hops away it is, and cascading does not insulate the primary from it", "source_quote": null, "source_url": null }, { "text": "Under remote_apply, a query that holds up replay on a synchronous replica holds up every commit on the primary waiting on it, for as long as max_standby_streaming_delay allows.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "a query that holds up replay on that replica holds up every commit on the primary that is waiting on it", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209556" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:57.666Z" } }, { "rank": 22, "id": "st_20261024_652", "title": "huggingface/diffusers v0.42.0", "url": "https://github.com/huggingface/diffusers/releases/tag/v0.42.0", "author": "sayakpaul", "site": "GitHub", "published_at": "2026-10-23T15:39:55.000Z", "first_published_at": "2026-10-23T15:39:55.000Z", "reading_minutes": 6, "depth": "claim_checked", "format": "release", "summary": "Hugging Face released Diffusers v0.42.0, adding Python 3.11+ requirement and support for several new models: FLUX.3 Image Klein, Bria Fibo 2.0, Kandinsky 6 video generation with audio, SANA-WM camera-controlled video, and RFD4-Proteína for protein design.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.711, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.715, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6483, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.63, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.73, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.54, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.5867, "spread": 0.08, "weight": 0.1 }, "practical_relevance": { "jev": 0.905, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8417, "spread": 0.105, "weight": 0.15 }, "substantiated": { "jev": 0.83, "judge_a": 0.6, "judge_b": 0.92, "mean": 0.7833, "spread": 0.32, "weight": 0.1 }, "technical_depth": { "jev": 0.765, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7483, "spread": 0.08, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 1, "mean": 0.9333, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.08, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.14, "judge_a": 0.15, "judge_b": 0.04, "mean": 0.11, "spread": 0.11, "weight": -0.15 } } }, "claims": [ { "text": "Diffusers v0.42.0 now requires Python 3.11 or newer, dropping Python 3.10 support.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "Python 3.10 support is dropped. Upgrade to Python 3.11 or newer before installing this release", "source_quote": null, "source_url": "https://github.com/huggingface/diffusers/pull/15030" }, { "text": "Bria Fibo 2.0's edit-turbo checkpoint runs in four steps without classifier-free guidance, and the model supports editing with up to five reference images.", "kind": "number", "label": "matches", "drift": null, "article_quote": "the edit-turbo checkpoint, which runs in four steps without classifier-free guidance", "source_quote": "briaai/Fibo-2.0-edit-turbo , distilled for editing in 4 steps without guidance (guidance_scale=1.0", "source_url": "https://huggingface.co/docs/diffusers/main/en/api/pipelines/bria_fibo2" }, { "text": "Kandinsky 6 generates video with synchronized audio from text or a reference image through Kandinsky6TI2VAPipeline, with a separate Kandinsky6SRPipeline for video super-resolution.", "kind": "version", "label": "matches", "drift": null, "article_quote": "Kandinsky 6 generates video with synchronized audio from text or a reference image through Kandinsky6TI2VAPipeline", "source_quote": "Kandinsky 6 is a family of video generation models from Kandinsky Lab . The main model generates video and synchronized audio from text or a reference image with a single…", "source_url": "https://huggingface.co/docs/diffusers/main/en/api/pipelines/kandinsky6" }, { "text": "The supported huggingface_hub range is now >=1.32.0,<3.0.", "kind": "version", "label": "matches", "drift": null, "article_quote": "the supported huggingface_hub range is now >=1.32.0,<3.0", "source_quote": "Bumps the `huggingface_hub` upper bound from `<2.0` to `<3.0` now that [huggingface_hub v2.0.0](https://github.com/huggingface/huggingface_hub/releases/tag/v2.0.0) is out", "source_url": "https://github.com/huggingface/diffusers/pull/14863" }, { "text": "The packed-latents fallback for FLUX, Krea 2 and LTX-Video modular pipelines is deprecated for removal in version 0.47.0.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "The fallback for packed latents is deprecated for removal in 0.47.0", "source_quote": null, "source_url": "https://github.com/huggingface/diffusers/pull/14737" }, { "text": "AutoencoderKLKVAEAudio adds encoding and decoding for 48 kHz audio.", "kind": "number", "label": "matches", "drift": null, "article_quote": "AutoencoderKLKVAEAudio adds encoding and decoding for 48 kHz audio", "source_quote": "compresses/reconstructs raw full-band (48 kHz) waveforms into continuous latents", "source_url": "https://github.com/huggingface/diffusers/pull/14384" }, { "text": "TorchTPU integration covers eager execution, torch.compile, and tensor parallelism, and tensor parallelism also gains Intel XPU support.", "kind": "version", "label": "couldnt_check", "drift": null, "article_quote": "TorchTPU integration covers eager execution, torch.compile, and tensor parallelism. Tensor parallelism also gains Intel XPU support", "source_quote": null, "source_url": "https://github.com/huggingface/diffusers/pull/14039" }, { "text": "Having a TorchAO version older than 0.15 installed no longer breaks model imports, though TorchAO quantization still requires 0.15 or newer.", "kind": "version", "label": "matches", "drift": null, "article_quote": "having a TorchAO version older than 0.15 installed no longer breaks model imports. TorchAO quantization still requires version 0.15 or newer", "source_quote": "The TorchAO quantizer already requires torchao >= 0.15 (`validate_environment`, and `TorchAoConfig` itself), so this only imports `FqnToConfig` and `quantize_` from that version on", "source_url": "https://github.com/huggingface/diffusers/pull/15095" } ], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:57.960Z" } }, { "rank": 23, "id": "st_20261024_265", "title": "Cutting our inference bill by 68% by being obsessive about optimization", "url": "https://blog.cloudflare.com/obsessive-inference-optimization/", "author": null, "site": "Cloudflare Blog", "published_at": "2026-10-23T07:03:47.000Z", "reading_minutes": 14, "depth": "three_judges", "format": "engineering_writeup", "summary": "Cloudflare's engineering team reports cutting the average cost of its internal AI code reviewer from $2.43 to $0.77 per review, a 68% reduction, while also halving median review time.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "established_project", "categories": [ "programming", "ai", "devops" ] }, "score": { "total": 0.7109, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.59, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5533, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.72, "judge_b": 0.94, "mean": 0.8167, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.483, "judge_a": 0.45, "judge_b": 0.63, "mean": 0.5211, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.898, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8525, "spread": 0.097, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.78, "judge_b": 0.86, "mean": 0.87, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.88, "judge_b": 0.91, "mean": 0.9258, "spread": 0.108, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9167, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.0967, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1833, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50205383" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:57.790Z" } }, { "rank": 24, "id": "st_20261024_946", "title": "php2shell: Millions of PHP services in Germany are vulnerable", "url": "https://www.heise.de/en/news/php2shell-Millions-of-PHP-services-in-Germany-are-vulnerable-11498549.html", "author": "heise online", "site": "heise online English", "published_at": "2026-10-23T12:30:00.000Z", "first_published_at": "2026-10-23T12:30:00.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "news_report", "summary": "Censys scans show tens of millions of internet-facing PHP services still running versions vulnerable to the \"php2shell\" flaws (CVE-2026-109618 and CVE-2026-109619, both rated critical), which allow remote code injection. Germany ranks second worldwide, with about 4.2 million potentially vulnerable services.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security", "devops", "it" ] }, "score": { "total": 0.7099, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.74, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.74, "spread": 0.08, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.84, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.68, "judge_a": 0.8, "judge_b": 0.72, "mean": 0.7333, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.823, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8408, "spread": 0.1, "weight": 0.15 }, "substantiated": { "jev": 0.697, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6589, "spread": 0.28, "weight": 0.1 }, "technical_depth": { "jev": 0.708, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5758, "spread": 0.308, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.92, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.03, "judge_b": 0.03, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.52, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.3467, "spread": 0.3, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:57.954Z" } }, { "rank": 25, "id": "st_20261024_720", "title": "Bun v1.4.4: now including `bun fmt`, a built-in formatter", "url": "https://bun.com/blog/bun-v1.4.4", "author": null, "site": "Bun", "published_at": "2026-10-23T14:05:19.000Z", "reading_minutes": 75, "depth": "claim_checked", "format": "release", "summary": "Bun v1.4.4 adds bun fmt, a built-in code formatter compatible with Prettier and oxfmt configs, supporting most common languages. Reported benchmarks show it running roughly 3x to 286x faster than Prettier or oxfmt on large repositories, with output matching Prettier on about 99% of its own test cases.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "devops" ] }, "score": { "total": 0.7092, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.763, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.6108, "spread": 0.413, "weight": 0.25 }, "reader_interest": { "jev": 0.8, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.81, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.543, "judge_a": 0.6, "judge_b": 0.63, "mean": 0.5911, "spread": 0.087, "weight": 0.1 }, "practical_relevance": { "jev": 0.96, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.86, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8556, "spread": 0.287, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.9, "judge_b": 0.9, "mean": 0.9317, "spread": 0.095, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9167, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.1567, "spread": 0.25, "weight": -0.15 }, "overclaim": { "jev": 0.2, "judge_a": 0.45, "judge_b": 0.16, "mean": 0.27, "spread": 0.29, "weight": -0.15 } } }, "claims": [ { "text": "In Prettier's own repository (2,616 files), bun fmt --check takes about 60 ms while Prettier 3.9.10 takes 17 seconds.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "In Prettier's own repository (2,616 files, with their config), bun fmt --check takes 60 ms and Prettier takes 17 seconds", "source_quote": null, "source_url": null }, { "text": "bun fmt is 3x to 6x faster than oxfmt 0.74.0 on projects that already use oxfmt with their own config.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "On projects that already use oxfmt, with their own config, it's 3x to 6x faster than oxfmt 0.74.0", "source_quote": null, "source_url": null }, { "text": "bun fmt's output was byte-for-byte identical to Prettier 3.9.9's on 9,648 of 9,731 Prettier formatting tests (99.1%).", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "Total 9,731 9,648 99.1%", "source_quote": null, "source_url": null }, { "text": "The JavaScript and TypeScript printer in bun fmt is a port of oxc_formatter, which is a port of Biome's formatter.", "kind": "generalization", "label": "no_source", "drift": null, "article_quote": "the JavaScript and TypeScript printer is a port of oxc_formatter, which is a port of Biome's formatter", "source_quote": null, "source_url": null }, { "text": "bun check is 1.4x to 7.8x faster on large projects in v1.4.4 than in v1.4.3, with MUI docs going from 10.28 s to 1.31 s.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "bun check is 1.4x to 7.8x faster on large projects", "source_quote": null, "source_url": null }, { "text": "Bun.SQL uses up to 48% less client-side CPU per query with PostgreSQL.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "Bun.SQL uses up to 48% less CPU per query", "source_quote": null, "source_url": null }, { "text": "In Bun v1.4.3, all five bun install retries happened within 5 ms, so a brief registry outage would fail the install.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "In Bun v1.4.3, all five retries happened within 5 ms, so a brief registry outage would fail the install", "source_quote": null, "source_url": null }, { "text": "The JavaScriptCore upgrade in this release includes 1,013 upstream WebKit commits and makes n.toLocaleString(\"en-US\") 54x faster over 100,000 calls.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "This release upgrades JavaScriptCore with 1,013 upstream WebKit commits", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209223" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:57.876Z" } }, { "rank": 26, "id": "st_20261024_1591", "title": "AgentPhishing: A Benchmark of LLM Email Agents Under Adaptive Attack", "url": "https://arxiv.org/abs/2610.31404", "author": "Justin Tan, Miao Li, Ding Zhao", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:07:49.000Z", "reading_minutes": 53, "depth": "three_judges", "format": "original_research", "summary": "AgentPhishing is a benchmark for testing whether LLM email agents detect phishing and whether they comply with malicious requests in messages, including emails optimized against each agent. Its PhishOpt framework refines email-generation skills using agent feedback and derives a universal skill from persuasion analysis.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.7089, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.655, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6617, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.7, "judge_b": 0.92, "mean": 0.7733, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.683, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6678, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.713, "judge_a": 0.6, "judge_b": 0.8, "mean": 0.7042, "spread": 0.2, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.55, "judge_b": 0.8, "mean": 0.7711, "spread": 0.413, "weight": 0.1 }, "technical_depth": { "jev": 0.945, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.8583, "spread": 0.245, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.35, "judge_b": 0.08, "mean": 0.2067, "spread": 0.27, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.075Z" } }, { "rank": 27, "id": "st_20261024_1522", "title": "Behind Aggregate Scores: What Systematically Varying Arena-Hard's LLM Judge Reveals", "url": "https://arxiv.org/abs/2610.30512", "author": "Louise Yorke", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T22:46:14.000Z", "reading_minutes": 43, "depth": "three_judges", "format": "original_research", "summary": "This workshop paper shows that LLM judges used in Arena-Hard can produce stable aggregate scores while individual prompt verdicts change substantially. Rerunning an unchanged GPT-4.1 setup altered 21.2% of prompt outcomes, and some instruction packages changed more.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.708, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.688, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6392, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.72, "judge_b": 0.93, "mean": 0.7767, "spread": 0.25, "weight": 0.2 }, "audience_breadth": { "jev": 0.617, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.6156, "spread": 0.13, "weight": 0.1 }, "practical_relevance": { "jev": 0.77, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7633, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.75, "judge_b": 0.78, "mean": 0.8411, "spread": 0.243, "weight": 0.1 }, "technical_depth": { "jev": 0.985, "judge_a": 0.7, "judge_b": 0.92, "mean": 0.8683, "spread": 0.285, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.97, "mean": 0.8633, "spread": 0.27, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2367, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.236Z" } }, { "rank": 28, "id": "st_20261024_1402", "title": "Can LLMs Improve Their Own Harness? A Controlled Study on ARC-AGI-3", "url": "https://arxiv.org/abs/2610.30499", "author": "Jens Lehmann, Sahar Vahdati", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T22:29:47.000Z", "reading_minutes": 49, "depth": "three_judges", "format": "original_research", "summary": "A study tests whether frontier language models can autonomously build and then revise the software harness that wraps them on ARC-AGI-3, an interactive benchmark of novel games, without human edits.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.7072, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.668, "judge_a": 0.55, "judge_b": 0.76, "mean": 0.6592, "spread": 0.21, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.79, "spread": 0.19, "weight": 0.2 }, "audience_breadth": { "jev": 0.567, "judge_a": 0.6, "judge_b": 0.58, "mean": 0.5822, "spread": 0.033, "weight": 0.1 }, "practical_relevance": { "jev": 0.605, "judge_a": 0.6, "judge_b": 0.68, "mean": 0.6283, "spread": 0.08, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8789, "spread": 0.177, "weight": 0.1 }, "technical_depth": { "jev": 0.95, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.8933, "spread": 0.15, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8967, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.2, "judge_b": 0.14, "mean": 0.1967, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.347Z" } }, { "rank": 29, "id": "st_20261024_105", "title": "Keycloak: Unauthenticated account takeover via reset-credentials flow bypass", "url": "https://access.redhat.com/security/cve/cve-2026-18963", "author": null, "site": "access.redhat.com", "published_at": "2026-10-23T03:54:50.000Z", "reading_minutes": 1, "depth": "three_judges", "format": "news_report", "summary": "A critical flaw in the reset-credentials flow of Keycloak, the identity management engine behind Red Hat Build of Keycloak, lets an unauthenticated remote attacker set new passwords for any user without completing email verification, enabling full account takeover with no user interaction (CVSS 9.1).", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security", "it" ] }, "score": { "total": 0.7062, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.703, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7075, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.71, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.8167, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.337, "judge_a": 0.6, "judge_b": 0.48, "mean": 0.4722, "spread": 0.263, "weight": 0.1 }, "practical_relevance": { "jev": 0.903, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8842, "spread": 0.052, "weight": 0.15 }, "substantiated": { "jev": 0.463, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.6544, "spread": 0.437, "weight": 0.1 }, "technical_depth": { "jev": 0.71, "judge_a": 0.4, "judge_b": 0.55, "mean": 0.5533, "spread": 0.31, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9333, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.2, "judge_b": 0.03, "mean": 0.15, "spread": 0.19, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204334" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:58.104Z" } }, { "rank": 30, "id": "st_20261024_1569", "title": "Prompt Sensitivity in Tool-Augmented Model Comparisons", "url": "https://arxiv.org/abs/2610.31148", "author": "Mateusz Modrzejewski, Tomasz Radzikowski", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T12:07:01.000Z", "reading_minutes": 54, "depth": "three_judges", "format": "original_research", "summary": "This paper studies whether giving language models tools makes pairwise benchmark comparisons more stable across different wordings of the same question. The authors separate the effects of the answer readout, reasoning, the decision to call a tool, and access to tool reports, using matched conditions on six open models across…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.704, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.655, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6083, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.7467, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.57, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.59, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.718, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6992, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8889, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.85, "judge_b": 0.93, "mean": 0.925, "spread": 0.145, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9033, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0233, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.25, "judge_b": 0.12, "mean": 0.1967, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.407Z" } }, { "rank": 31, "id": "st_20261024_1413", "title": "Scaling Laws for Safety Probes", "url": "https://arxiv.org/abs/2610.30678", "author": "Yuchen Wu (University of Washington), Jerry Wei (Anthropic)", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T03:43:33.000Z", "reading_minutes": 107, "depth": "three_judges", "format": "original_research", "summary": "This paper studies whether the performance of safety probes, small classifiers trained on language models' internal activations to monitor for harms, can be predicted in advance.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.7036, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.665, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.665, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.91, "mean": 0.7167, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.513, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5444, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.68, "judge_a": 0.6, "judge_b": 0.79, "mean": 0.69, "spread": 0.19, "weight": 0.15 }, "substantiated": { "jev": 0.973, "judge_a": 0.85, "judge_b": 0.93, "mean": 0.9178, "spread": 0.123, "weight": 0.1 }, "technical_depth": { "jev": 0.978, "judge_a": 0.85, "judge_b": 0.97, "mean": 0.9325, "spread": 0.128, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.89, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0367, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.2, "judge_b": 0.14, "mean": 0.2167, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.400Z" } }, { "rank": 32, "id": "st_20261024_1514", "title": "High-Value Training Data from Noisy, Diverse Sources", "url": "https://arxiv.org/abs/2610.30421", "author": "Hedi Driss, Evan Frick, Anastasios N. Angelopoulos et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T20:32:43.000Z", "reading_minutes": 56, "depth": "three_judges", "format": "original_research", "summary": "This paper proposes a scalable method for learning per-example training weights by using metagradients through the training process, optimizing weights directly for performance on a small, high-quality target dataset. It extends this with parametric weighting rules that generalize to new data and transfer from small to larger models.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "data" ] }, "score": { "total": 0.7027, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.708, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6625, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.72, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.673, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6644, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.693, "judge_a": 0.6, "judge_b": 0.8, "mean": 0.6975, "spread": 0.2, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8744, "spread": 0.163, "weight": 0.1 }, "technical_depth": { "jev": 0.928, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.9092, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.8567, "spread": 0.25, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.25, "judge_b": 0.2, "mean": 0.2367, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.549Z" } }, { "rank": 33, "id": "st_20261024_36", "title": "Humming: Fast, Flexible Kernels for Quantized Inference", "url": "https://vllm.ai/blog/2026-10-21-humming", "author": "vLLM Team", "site": "vllm.ai", "published_at": "2026-10-23T01:42:24.000Z", "first_published_at": "2026-10-21T00:00:00.000Z", "reading_minutes": 15, "depth": "three_judges", "format": "release", "summary": "Ant Group and the vLLM community have open-sourced Humming, a general-purpose kernel library for quantized inference. It supports NVIDIA GPUs from Turing onward plus some CUDA-compatible devices, handles many weight and activation formats for dense and MoE models, and reports better performance than vLLM Marlin, FlashInfer, and DeepGEMM on many…", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai", "hardware" ] }, "score": { "total": 0.7022, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.718, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6492, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.78, "judge_a": 0.72, "judge_b": 0.9, "mean": 0.8, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.577, "judge_a": 0.55, "judge_b": 0.7, "mean": 0.6089, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.888, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8558, "spread": 0.087, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8056, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 0.913, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.9042, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9367, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.1667, "spread": 0.18, "weight": -0.15 }, "overclaim": { "jev": 0.3, "judge_a": 0.4, "judge_b": 0.28, "mean": 0.3267, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203547" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:53:58.478Z" } }, { "rank": 34, "id": "st_20261024_1159", "title": "Writing Code vs. Shipping Code: Productivity Across Generations of AI Code Tools", "url": "https://www.nber.org/papers/w35275", "author": "Mert Demirer,", "site": "NBER", "published_at": "2026-10-23T18:16:14.000Z", "reading_minutes": 1, "depth": "three_judges", "format": "original_research", "summary": "A study of over 500,000 GitHub developers, matched to their AI usage telemetry, finds that autocomplete, interactive coding agents, and autonomous coding agents raise commit activity by 30%, 180%, and 240% respectively.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.7018, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.725, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.685, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.82, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.853, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8044, "spread": 0.16, "weight": 0.1 }, "practical_relevance": { "jev": 0.623, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6675, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.857, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.7389, "spread": 0.257, "weight": 0.1 }, "technical_depth": { "jev": 0.653, "judge_a": 0.45, "judge_b": 0.74, "mean": 0.6142, "spread": 0.29, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.97, "mean": 0.8667, "spread": 0.27, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213207" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:58.526Z" } }, { "rank": 35, "id": "st_20261024_1419", "title": "Is Complex Agentic Memory Always Needed? Naive RAG Is Stronger Than We Thought", "url": "https://arxiv.org/abs/2610.30755", "author": "Zhengye Bao, Shengtao Zhang, Haokai Zhang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T05:23:16.000Z", "reading_minutes": 89, "depth": "three_judges", "format": "original_research", "summary": "A study compares Raw-RAG, a naive retrieval baseline that uses dense retrieval and reranking over raw conversation history, against complex agentic memory systems that mutate, restructure, or learn retrieval strategies.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.7005, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.693, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6742, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.7, "judge_b": 0.92, "mean": 0.7667, "spread": 0.24, "weight": 0.2 }, "audience_breadth": { "jev": 0.587, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6022, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.78, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7667, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.8544, "spread": 0.203, "weight": 0.1 }, "technical_depth": { "jev": 0.93, "judge_a": 0.75, "judge_b": 0.88, "mean": 0.8533, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8667, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.48, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.32, "spread": 0.3, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.683Z" } }, { "rank": 36, "id": "st_20261024_837", "title": "Preliminary Report Oct 21st Security Incident", "url": "https://blog.railway.com/p/preliminary-report-october-security-incident", "author": "Jacob Cooper & Angelo Saraceno", "site": "Railway Blog", "published_at": "2026-10-23T16:48:50.000Z", "reading_minutes": 9, "depth": "three_judges", "format": "news_report", "summary": "Railway, a cloud deployment platform, disclosed a preliminary report on an October 2026 security incident in which an attacker, targeting AI provider API keys, accessed internal systems and decrypted some customer environment variables.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai", "security", "devops" ] }, "score": { "total": 0.7001, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.68, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.8, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.8633, "spread": 0.14, "weight": 0.2 }, "audience_breadth": { "jev": 0.377, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.5989, "spread": 0.343, "weight": 0.1 }, "practical_relevance": { "jev": 0.918, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8658, "spread": 0.117, "weight": 0.15 }, "substantiated": { "jev": 0.623, "judge_a": 0.5, "judge_b": 0.82, "mean": 0.6478, "spread": 0.32, "weight": 0.1 }, "technical_depth": { "jev": 0.743, "judge_a": 0.45, "judge_b": 0.7, "mean": 0.6308, "spread": 0.293, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9333, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.1033, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.2, "judge_b": 0.35, "mean": 0.2533, "spread": 0.15, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 36, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211898" }, "lobsters": null, "line": "Elsewhere: HN 36 pts, not on front page", "checked_at": "2026-10-24T08:53:58.604Z" } }, { "rank": 37, "id": "st_20261024_856", "title": "OlmoEarth v1.3: embeddings for the whole planet", "url": "https://huggingface.co/blog/allenai/olmoearth-v1-3", "author": null, "site": "huggingface.co", "published_at": "2026-10-23T16:34:24.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "release", "summary": "Allen Institute for AI has released OlmoEarth v1.3, a satellite embedding model, along with precomputed annual 10-meter embeddings for the entire planet and a browser-based Embeddings Explorer that lets users map and classify landscapes without GPUs or sign-in.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "ai", "data" ] }, "score": { "total": 0.6999, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.715, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7117, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.79, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.437, "judge_a": 0.55, "judge_b": 0.62, "mean": 0.5356, "spread": 0.183, "weight": 0.1 }, "practical_relevance": { "jev": 0.855, "judge_a": 0.7, "judge_b": 0.8, "mean": 0.785, "spread": 0.155, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8956, "spread": 0.177, "weight": 0.1 }, "technical_depth": { "jev": 0.758, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.7658, "spread": 0.14, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.94, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.1867, "spread": 0.22, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.3, "judge_b": 0.2, "mean": 0.2633, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211654" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:58.652Z" } }, { "rank": 38, "id": "st_20261024_1629", "title": "Invisibilization as a Source of Probability Bias in Language Models", "url": "https://arxiv.org/abs/2610.30785", "author": "Christian Arnold, Ernesto Calvo, Guadalupe Gonzalez et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T06:03:35.000Z", "reading_minutes": 43, "depth": "three_judges", "format": "original_research", "summary": "This paper measures how leading language models from OpenAI and Anthropic often refuse to assess unverifiable reports when the reporter is described by race, ethnicity, or other protected attributes, rather than weighing the evidence.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6974, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.733, "judge_a": 0.5, "judge_b": 0.82, "mean": 0.6842, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.7767, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.74, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6867, "spread": 0.14, "weight": 0.1 }, "practical_relevance": { "jev": 0.763, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7142, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8322, "spread": 0.277, "weight": 0.1 }, "technical_depth": { "jev": 0.965, "judge_a": 0.6, "judge_b": 0.93, "mean": 0.8317, "spread": 0.365, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8933, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.32, "judge_a": 0.4, "judge_b": 0.38, "mean": 0.3667, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.853Z" } }, { "rank": 39, "id": "st_20261024_1564", "title": "What Pairwise Evaluation Misses: Reward Scores across Prompts", "url": "https://arxiv.org/abs/2610.31090", "author": "Mateusz Modrzejewski, Tomasz Radzikowski", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T11:00:10.000Z", "reading_minutes": 51, "depth": "three_judges", "format": "original_research", "summary": "This paper proposes a human-referenced audit of reward models, testing how much their scores vary between prompts compared with human ratings of the same responses.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6966, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.68, "judge_a": 0.55, "judge_b": 0.74, "mean": 0.6567, "spread": 0.19, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.6967, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.51, "judge_a": 0.55, "judge_b": 0.52, "mean": 0.5267, "spread": 0.04, "weight": 0.1 }, "practical_relevance": { "jev": 0.693, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6708, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8989, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.9275, "spread": 0.143, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8967, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0267, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.25, "judge_b": 0.08, "mean": 0.19, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.887Z" } }, { "rank": 40, "id": "st_20261024_1609", "title": "Low-Bit N-Gram Embeddings for Language Models", "url": "https://arxiv.org/abs/2610.31646", "author": "Stephen Chung, Wenyu Du", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:56:13.000Z", "reading_minutes": 31, "depth": "three_judges", "format": "original_research", "summary": "The paper proposes storing n-gram embedding tables, which some open-weight language models use to boost performance, at low bit precision to cut their memory footprint.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6935, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.618, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5958, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.7567, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.527, "judge_a": 0.5, "judge_b": 0.58, "mean": 0.5356, "spread": 0.08, "weight": 0.1 }, "practical_relevance": { "jev": 0.76, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7467, "spread": 0.08, "weight": 0.15 }, "substantiated": { "jev": 0.967, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8889, "spread": 0.167, "weight": 0.1 }, "technical_depth": { "jev": 0.983, "judge_a": 0.85, "judge_b": 0.96, "mean": 0.9308, "spread": 0.133, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8967, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.0467, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.32, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2467, "spread": 0.2, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:58.991Z" } }, { "rank": 41, "id": "st_20261024_1526", "title": "Lifetime-valid decisions by continually evolving AI analysts", "url": "https://arxiv.org/abs/2610.30562", "author": "Xiaojing Sun, Yuhan Zeng, Xiao Wang", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T00:34:06.000Z", "reading_minutes": 95, "depth": "three_judges", "format": "original_research", "summary": "This paper addresses a gap in Type I error control for AI analysts that keep changing: prompt edits, tool additions, and model swaps can each shift the false-positive rate, and adaptive updates can overfit to past feedback.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "data" ] }, "score": { "total": 0.6918, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.68, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6533, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7033, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.7, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.64, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.695, "judge_a": 0.5, "judge_b": 0.8, "mean": 0.665, "spread": 0.3, "weight": 0.15 }, "substantiated": { "jev": 0.943, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.8911, "spread": 0.143, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.9158, "spread": 0.188, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8933, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.42, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.2667, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.059Z" } }, { "rank": 42, "id": "st_20261024_200", "title": "Whistle speech-to-text running offline on a 16mb ESP32", "url": "https://github.com/andrisgauracs/whistle-esp32", "author": null, "site": "GitHub", "published_at": "2026-10-23T06:45:29.000Z", "reading_minutes": 8, "depth": "three_judges", "format": "engineering_writeup", "summary": "A maker project runs Whistle, a speech-to-text model from Cactus Compute, entirely offline on a $10 ESP32-S3 microcontroller, transcribing spoken commands in seven languages to control an LED.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "new_side_project", "categories": [ "programming", "ai", "hardware" ] }, "score": { "total": 0.6917, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.52, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5467, "spread": 0.12, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.75, "judge_b": 0.94, "mean": 0.81, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.35, "judge_a": 0.5, "judge_b": 0.42, "mean": 0.4233, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.813, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7308, "spread": 0.213, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.75, "judge_b": 0.91, "mean": 0.8822, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 1, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.96, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9333, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.0567, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1867, "spread": 0.2, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50205260" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:58.981Z" } }, { "rank": 43, "id": "st_20261024_1642", "title": "Padding to k: A Minimum Count in the Schema Makes Open-Weight Models Pad Their Extraction Lists", "url": "https://arxiv.org/abs/2610.30997", "author": "Abhishek Barali, Aayush Paudel", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:38:58.000Z", "depth": "three_judges", "format": "original_research", "summary": "When a JSON schema sets a minimum number of list entries but the document contains fewer items, open-weight language models tend to pad their output to meet the count.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6906, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.645, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5883, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7233, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.537, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.5856, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.818, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7925, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.8744, "spread": 0.293, "weight": 0.1 }, "technical_depth": { "jev": 0.75, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7233, "spread": 0.22, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9067, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.2, "judge_b": 0.05, "mean": 0.1533, "spread": 0.16, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.187Z" } }, { "rank": 44, "id": "st_20261024_1494", "title": "Agentic Verified Optimization of Real-World Software", "url": "https://arxiv.org/abs/2610.30188", "author": "Zhe Ye, Udaya Ghai, Zhizhen Qin et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-20T17:33:22.000Z", "reading_minutes": 59, "depth": "three_judges", "format": "original_research", "summary": "Hrih News: A research paper proposes \"agentic verified optimization,\" where a coding agent speeds up an existing program and submits a machine-checked Lean 4 proof that the faster version matches the original on all inputs, rather than relying on tests that agents can game.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6888, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.658, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6925, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.72, "judge_b": 0.92, "mean": 0.7533, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.633, "judge_a": 0.5, "judge_b": 0.67, "mean": 0.6011, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.513, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.5808, "spread": 0.33, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8478, "spread": 0.263, "weight": 0.1 }, "technical_depth": { "jev": 0.925, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.9083, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8567, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0433, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.2467, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.267Z" } }, { "rank": 45, "id": "st_20261024_1748", "title": "The Value of System 1 Models in Medicine", "url": "https://arxiv.org/abs/2610.31315", "author": "Xiufeng Yang, Konstantinos Vrettos, Zeineb Ben Chaaben et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:08:52.000Z", "depth": "three_judges", "format": "original_research", "summary": "A study tested four \"System 1\" AI models, which answer clinical questions instantly with a probability score rather than generating free text, across 86 public medical tasks in 20 languages.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6885, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.66, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.6767, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.6, "judge_b": 0.93, "mean": 0.7267, "spread": 0.33, "weight": 0.2 }, "audience_breadth": { "jev": 0.563, "judge_a": 0.4, "judge_b": 0.82, "mean": 0.5944, "spread": 0.42, "weight": 0.1 }, "practical_relevance": { "jev": 0.673, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6675, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8856, "spread": 0.147, "weight": 0.1 }, "technical_depth": { "jev": 0.715, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.725, "spread": 0.26, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.97, "mean": 0.8933, "spread": 0.17, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.2, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.277Z" } }, { "rank": 46, "id": "st_20261024_500", "title": "Review of \"Misalignment in high-stakes settings\" in Anthropic's Risk Report", "url": "https://metr.org/blog/2026-10-20-misalignment-risk-report-august-2026-review/", "author": null, "site": "metr.org", "published_at": "2026-10-23T12:50:40.000Z", "first_published_at": "2026-10-20T00:00:00.000Z", "reading_minutes": 26, "depth": "three_judges", "format": "analysis_opinion", "summary": "METR's external review of the \"Misalignment in high-stakes settings\" section of Anthropic's August 2026 Risk Report agrees that catastrophic misalignment risk from covered models is low, but attributes this mainly to limited model capabilities rather than safeguards.", "also_covered_by": [ { "site": "anthropic.com", "url": "https://www.anthropic.com/research/alignment-assessment-ukaisi", "title": "An alignment assessment of the UK AISI cybertesting incident" } ], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.688, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.653, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.6742, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.79, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.43, "judge_a": 0.55, "judge_b": 0.7, "mean": 0.56, "spread": 0.27, "weight": 0.1 }, "practical_relevance": { "jev": 0.538, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6192, "spread": 0.183, "weight": 0.15 }, "substantiated": { "jev": 0.757, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.7789, "spread": 0.043, "weight": 0.1 }, "technical_depth": { "jev": 0.84, "judge_a": 0.75, "judge_b": 0.78, "mean": 0.79, "spread": 0.09, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9067, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.02, "mean": 0.0333, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.2, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208187" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:59.194Z" } }, { "rank": 47, "id": "st_20261024_870", "title": "GitHub to Forgejo: what to get right before a one-afternoon switch", "url": "https://amaiko.ai/engineering/migrating-from-github-to-forgejo", "author": "Florian Schauer", "site": "amaiko.ai", "published_at": "2026-10-23T16:22:11.000Z", "reading_minutes": 5, "depth": "claim_checked", "format": "engineering_writeup", "summary": "A company moved its production-deploy repositories, CI and container registry from GitHub to a self-hosted Forgejo instance after two GitHub Actions outages in August 2026 blocked its releases.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "devops" ] }, "score": { "total": 0.6865, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.51, "judge_a": 0.3, "judge_b": 0.58, "mean": 0.4633, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.8, "judge_a": 0.82, "judge_b": 0.82, "mean": 0.8133, "spread": 0.02, "weight": 0.2 }, "audience_breadth": { "jev": 0.407, "judge_a": 0.5, "judge_b": 0.42, "mean": 0.4422, "spread": 0.093, "weight": 0.1 }, "practical_relevance": { "jev": 0.918, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8592, "spread": 0.117, "weight": 0.15 }, "substantiated": { "jev": 0.863, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7811, "spread": 0.163, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.85, "judge_b": 0.86, "mean": 0.9025, "spread": 0.148, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.9, "spread": 0.16, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0367, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.13, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.12, "spread": 0.07, "weight": -0.15 } } }, "claims": [ { "text": "At the peak of GitHub's Actions incident on 6 August 2026, 71% of runs failed.", "kind": "number", "label": "matches", "drift": null, "article_quote": "At the peak, 71% of runs failed", "source_quote": "At peak, 71% of workflow runs saw infrastructure failures , and 75% of the remainder were delayed by more than 5 minutes", "source_url": "https://blog.incidenthub.cloud/github-actions-pages-outage-aug-6-2026" }, { "text": "GitHub's Actions incident on 6 August 2026 ran from 15:22 UTC until 02:04 the next morning.", "kind": "date", "label": "matches", "drift": null, "article_quote": "GitHub’s Actions incident ran from 15:22 UTC until 02:04 the next morning", "source_quote": "Impact window (status page): Aug 6 15:22 UTC to Aug 7 02:04 UTC (about 10 h 42 m from first investigating post to resolve", "source_url": "https://blog.incidenthub.cloud/github-actions-pages-outage-aug-6-2026" }, { "text": "GitHub had another outage on 17 August from 13:28 to 21:15 UTC.", "kind": "date", "label": "matches", "drift": null, "article_quote": "On 17 August, GitHub had another outage from 13:28 to 21:15 UTC", "source_quote": "On August 17, 2026, from 13:28–21:15 UTC (7h 47m), GitHub.com experienced elevated errors and latency across Issues, Pull Requests, APIs, Actions, and Copilot", "source_url": "https://github.com/orgs/community/discussions/205164" }, { "text": "The 17 August outage was GitHub's second significant incident that month.", "kind": "number", "label": "matches", "drift": null, "article_quote": "its second significant incident that month", "source_quote": "This was our second significant incident in August, following an actions failure on August 6", "source_url": "https://github.blog/news-insights/company-news/the-august-17-outage-and-the-work-ahead" }, { "text": "A self-hosted forge plus the team's own CI machines costs less than half of what GitHub did.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "a self-hosted forge plus our own CI machines costs less than half of what GitHub did", "source_quote": null, "source_url": null }, { "text": "The cutover to Forgejo took under two hours and the migration two days, after roughly two weeks of preparation.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "The cutover took under two hours, the migration two days, after roughly two weeks of preparation", "source_quote": null, "source_url": null }, { "text": "On a Forgejo runner, the runner overwrites job-level FORGEJO_TOKEN and GITHUB_TOKEN env values with the automatic job token, silently replacing a PAT.", "kind": "causal", "label": "no_source", "drift": null, "article_quote": "The runner applies workflow and job env, then overwrites both names with the automatic token, and only then applies step-level env", "source_quote": null, "source_url": null }, { "text": "ArgoCD 3.5 decides whether it has already tried an automated sync by comparing the current source to the last synced one, so changing every source resets that guard.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "ArgoCD 3.5 decides whether it has “already tried” an automated sync by comparing the current source to the last synced one", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211481" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:59.314Z" } }, { "rank": 48, "id": "st_20261024_287", "title": "What a Terrible Day to Have a Computer", "url": "https://blog.ricci.io/what-a-terrible-day-to-have-a-computer-d6b323r", "author": null, "site": "pckt", "published_at": "2026-10-23T08:53:39.000Z", "reading_minutes": 21, "depth": "three_judges", "format": "analysis_opinion", "summary": "A professor who administers research testbeds describes a day of scrambling to patch PHP, a language underpinning much of the web, after a critical remote code execution flaw was disclosed and an LLM-assisted exploit was expected to spread quickly.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai", "security", "devops", "it" ] }, "score": { "total": 0.6856, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.685, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6883, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.9, "judge_b": 0.9, "mean": 0.8333, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.947, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.8422, "spread": 0.167, "weight": 0.1 }, "practical_relevance": { "jev": 0.6, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7067, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.17, "judge_a": 0.6, "judge_b": 0.58, "mean": 0.45, "spread": 0.43, "weight": 0.1 }, "technical_depth": { "jev": 0.778, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7192, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.95, "mean": 0.8267, "spread": 0.35, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.02, "judge_b": 0.01, "mean": 0.03, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.25, "judge_b": 0.28, "mean": 0.2567, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206164" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, not on front page", "checked_at": "2026-10-24T08:53:59.319Z" } }, { "rank": 49, "id": "st_20261024_1566", "title": "ProductWebBench: Evaluating Website Continuity Beyond Task Completion", "url": "https://arxiv.org/abs/2610.31118", "author": "Jiajun Wu, Jian Yang, Yaxin Du et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T11:33:00.000Z", "reading_minutes": 59, "depth": "three_judges", "format": "original_research", "summary": "ProductWebBench is a benchmark that tests whether coding agents can finish website changes without breaking the rest of the site, not just whether the requested change appears.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6852, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.668, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6158, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.7067, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.59, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5533, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.68, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.68, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8911, "spread": 0.193, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.91, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.87, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0367, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.25, "judge_b": 0.12, "mean": 0.1933, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.566Z" } }, { "rank": 50, "id": "st_20261024_632", "title": "TanStack AI 1.0: Still Your AI, Still Your Way", "url": "https://tanstack.com/blog/tanstack-ai-1-0", "author": "co-authored by Alem Tuzlak, Jack Herrington, Tom Beckenham, and Tanner Linsley", "site": "TanStack", "published_at": "2026-10-23T13:10:36.000Z", "reading_minutes": 9, "depth": "three_judges", "format": "release", "summary": "TanStack AI has reached 1.0, with stable core APIs for chat, tools, media, MCP, persistence, and sandboxes, plus an experimental agent harness. It now supports 23 provider adapters and eight UI frameworks, and the team says it removed every API marked deprecated in 0.x.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6827, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.71, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.67, "spread": 0.11, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.81, "spread": 0.09, "weight": 0.2 }, "audience_breadth": { "jev": 0.47, "judge_a": 0.7, "judge_b": 0.62, "mean": 0.5967, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.97, "judge_a": 0.8, "judge_b": 0.8, "mean": 0.8567, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.803, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7078, "spread": 0.203, "weight": 0.1 }, "technical_depth": { "jev": 0.813, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7642, "spread": 0.113, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9333, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.35, "judge_b": 0.42, "mean": 0.28, "spread": 0.35, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.3, "judge_b": 0.2, "mean": 0.2233, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208433" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:59.405Z" } }, { "rank": 51, "id": "st_20261024_1483", "title": "Simulated Control Evals", "url": "https://arxiv.org/abs/2610.31593", "author": "Jeffrey Olmo, David Zhang, Tyler Tracy", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:26:42.000Z", "reading_minutes": 41, "depth": "three_judges", "format": "original_research", "summary": "Researchers propose simulated control evaluations (SCEs), which use a language model to generate tool responses in place of a live environment so that AI safety safeguards can be tested against an attacking agent using a deployment's own tasks and codebase.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6822, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.668, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6658, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.74, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.56, "judge_a": 0.4, "judge_b": 0.55, "mean": 0.5033, "spread": 0.16, "weight": 0.1 }, "practical_relevance": { "jev": 0.58, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.5933, "spread": 0.2, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8856, "spread": 0.177, "weight": 0.1 }, "technical_depth": { "jev": 0.96, "judge_a": 0.85, "judge_b": 0.97, "mean": 0.9267, "spread": 0.12, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8967, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.2467, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.594Z" } }, { "rank": 52, "id": "st_20261024_935", "title": "Reports of PHP bug exploitation attempts grow as experts downplay viability of exploits", "url": "https://therecord.media/php-php2shell-exploitation-attempts-pocs", "author": null, "site": "The Record (Recorded Future News)", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-23T04:00:00.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "news_report", "summary": "Reports of exploitation attempts against two PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619 (dubbed PHP2Shell), are multiplying, with Amazon citing activity dating back to October 17 and Arctic Wolf and Cobalt linking attempts to botnets such as Potassium and Mirai.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security" ] }, "score": { "total": 0.6814, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.715, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6617, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.8333, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.66, "mean": 0.64, "spread": 0.06, "weight": 0.1 }, "practical_relevance": { "jev": 0.7, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7267, "spread": 0.08, "weight": 0.15 }, "substantiated": { "jev": 0.523, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6478, "spread": 0.297, "weight": 0.1 }, "technical_depth": { "jev": 0.628, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5992, "spread": 0.27, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8667, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.2033, "spread": 0.25, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.658Z" } }, { "rank": 53, "id": "st_20261024_604", "title": "How to Speed Up VLA Training Step by Step:3× Acceleration on Pi0.5", "url": "https://baidu-baige.github.io/LoongForge/blog/2026-10-loongforge-pi05-3x-training-acceleration.html", "author": null, "site": "baidu-baige.github.io", "published_at": "2026-10-23T13:28:01.000Z", "reading_minutes": 12, "depth": "three_judges", "format": "engineering_writeup", "summary": "LoongForge reports a 3x training speedup for Pi0.5 supervised fine-tuning, reaching 24.0 samples/s in bf16 and 27.34 samples/s in FP8. The gains come from pruning 256 padding tokens for an unused camera slot, partially compiling the fixed-shape backbone with torch.compile, using FlexAttention to cut activation memory, recomputing only the vision…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "established_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6804, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.55, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5567, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.72, "judge_b": 0.9, "mean": 0.8033, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.377, "judge_a": 0.4, "judge_b": 0.5, "mean": 0.4256, "spread": 0.123, "weight": 0.1 }, "practical_relevance": { "jev": 0.835, "judge_a": 0.72, "judge_b": 0.78, "mean": 0.7783, "spread": 0.115, "weight": 0.15 }, "substantiated": { "jev": 0.897, "judge_a": 0.72, "judge_b": 0.86, "mean": 0.8256, "spread": 0.177, "weight": 0.1 }, "technical_depth": { "jev": 1, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.96, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9067, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1233, "spread": 0.15, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.1967, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208665" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:53:59.535Z" } }, { "rank": 54, "id": "st_20261024_1332", "title": "I ran the full 2.8T-parameter Kimi-K3 on a 64GB Mac Mini", "url": "https://github.com/sjdthree/kimi-k3-mlx-macmini", "author": null, "site": "GitHub", "published_at": "2026-10-23T22:19:23.000Z", "reading_minutes": 10, "depth": "three_judges", "format": "engineering_writeup", "summary": "A developer ran the complete 2.8-trillion-parameter Kimi-K3 checkpoint on a 64 GB M4 Pro Mac mini by streaming mixture-of-experts weights from an external SSD through an MLX and Metal implementation.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "new_side_project", "categories": [ "programming", "ai", "hardware" ] }, "score": { "total": 0.6804, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.565, "judge_a": 0.62, "judge_b": 0.72, "mean": 0.635, "spread": 0.155, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.8267, "spread": 0.25, "weight": 0.2 }, "audience_breadth": { "jev": 0.493, "judge_a": 0.6, "judge_b": 0.54, "mean": 0.5444, "spread": 0.107, "weight": 0.1 }, "practical_relevance": { "jev": 0.725, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6317, "spread": 0.275, "weight": 0.15 }, "substantiated": { "jev": 0.96, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.8067, "spread": 0.36, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.9108, "spread": 0.193, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9133, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.15, "judge_b": 0.03, "mean": 0.0867, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2867, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": false, "url": "https://news.ycombinator.com/item?id=50216152" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, not on front page", "checked_at": "2026-10-24T08:53:59.673Z" } }, { "rank": 55, "id": "st_20261024_1473", "title": "DIAL: Steerable Language Models with Distilled Instruction-Adherence Levels", "url": "https://arxiv.org/abs/2610.31438", "author": "Thanh Q. Tran, John Russell Himawan, Arun Verma et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:34:39.000Z", "reading_minutes": 100, "depth": "three_judges", "format": "original_research", "summary": "DIAL is a method for steering language models with fine-grained numerical control over how closely they follow general text instructions, such as response length or persona.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6766, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.645, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6083, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.5, "judge_b": 0.94, "mean": 0.6933, "spread": 0.44, "weight": 0.2 }, "audience_breadth": { "jev": 0.653, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5911, "spread": 0.32, "weight": 0.1 }, "practical_relevance": { "jev": 0.77, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6667, "spread": 0.33, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.8567, "spread": 0.27, "weight": 0.1 }, "technical_depth": { "jev": 0.968, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9158, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.93, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.01, "mean": 0.0533, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.2367, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:53:59.935Z" } }, { "rank": 56, "id": "st_20261024_120", "title": "The Customer Is the Builder: how AI coding agents change software economics", "url": "https://arxiv.org/abs/2610.30307", "author": null, "site": "arXiv", "published_at": "2026-10-23T03:24:24.000Z", "first_published_at": "2026-10-21T18:20:11.000Z", "reading_minutes": 52, "depth": "three_judges", "format": "original_research", "summary": "The paper argues that AI coding agents could shift the make-or-buy calculus for software, letting organizations and professional buyers build the functions they need instead of purchasing them.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6755, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.7, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6567, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.7867, "spread": 0.24, "weight": 0.2 }, "audience_breadth": { "jev": 0.887, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7889, "spread": 0.187, "weight": 0.1 }, "practical_relevance": { "jev": 0.585, "judge_a": 0.6, "judge_b": 0.68, "mean": 0.6217, "spread": 0.095, "weight": 0.15 }, "substantiated": { "jev": 0.94, "judge_a": 0.45, "judge_b": 0.82, "mean": 0.7367, "spread": 0.49, "weight": 0.1 }, "technical_depth": { "jev": 0.905, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.755, "spread": 0.405, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.6, "judge_b": 0.99, "mean": 0.8267, "spread": 0.39, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.1, "judge_b": 0.02, "mean": 0.0633, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.39, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.27, "spread": 0.27, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204149" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:53:59.748Z" } }, { "rank": 57, "id": "st_20261024_1664", "title": "Protection Is Not One Event: A Survey of RAG Protection Research by Measured Endpoint", "url": "https://arxiv.org/abs/2610.30193", "author": "Peiyang Liu, Xi Wang, Di Liang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T06:17:23.000Z", "depth": "three_judges", "format": "original_research", "summary": "A systematic survey of 118 full-text-confirmed RAG defense papers, coded by measured endpoint rather than claimed goal, finds that protection studies rarely evaluate comparable outcomes.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6753, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.698, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6258, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.73, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.553, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.5844, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.738, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7058, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8556, "spread": 0.287, "weight": 0.1 }, "technical_depth": { "jev": 0.758, "judge_a": 0.6, "judge_b": 0.8, "mean": 0.7192, "spread": 0.2, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.86, "spread": 0.25, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.0433, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.19, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:00.297Z" } }, { "rank": 58, "id": "st_20261024_1442", "title": "CANNBot-CAKE: Collaborative Agentic Kernel Evolution for Ascend NPU", "url": "https://arxiv.org/abs/2610.31034", "author": "Deng Pan, Xue-Jian Gao, Junyi Fan et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T10:12:45.000Z", "reading_minutes": 44, "depth": "three_judges", "format": "original_research", "summary": "CAKE is an agent system that generates and iteratively optimizes custom AscendC kernels for Huawei's Ascend NPUs, using multiple DSLs and reusable reviewed knowledge. On 53 benchmark operators it passed 100% of standard cases, 98.28% of hidden ones, and averaged 1.334x speedup, suggesting automation could ease Ascend's kernel-development bottleneck.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "product_release", "categories": [ "programming", "ai", "hardware" ] }, "score": { "total": 0.675, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.565, "judge_a": null, "judge_b": 0.76, "mean": 0.6625, "spread": 0.195, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": null, "judge_b": 0.88, "mean": 0.765, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.25, "judge_a": null, "judge_b": 0.48, "mean": 0.365, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.618, "judge_a": null, "judge_b": 0.78, "mean": 0.6988, "spread": 0.162, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": null, "judge_b": 0.8, "mean": 0.8733, "spread": 0.147, "weight": 0.1 }, "technical_depth": { "jev": 0.98, "judge_a": null, "judge_b": 0.94, "mean": 0.96, "spread": 0.04, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": null, "judge_b": 1, "mean": 0.955, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": null, "judge_b": 0.18, "mean": 0.13, "spread": 0.1, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": null, "judge_b": 0.28, "mean": 0.295, "spread": 0.03, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:00.277Z" } }, { "rank": 59, "id": "st_20261024_392", "title": "EL2 hypervisor and toolchain that boots macOS on the Snapdragon X2 Elite", "url": "https://github.com/Snapintosh/Snapintosh", "author": null, "site": "GitHub", "published_at": "2026-10-23T10:53:15.000Z", "reading_minutes": 6, "depth": "claim_checked", "format": "engineering_writeup", "summary": "Snapintosh is an experimental project that boots Apple's stock macOS 27 kernel on an ASUS Zenbook A16 with a Snapdragon X2 Elite, using a custom EL2 hypervisor, a UEFI proxy, and a homegrown Adreno GPU driver stack with its own Metal plug-in and shader compiler.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "early_project", "categories": [ "programming", "hardware" ] }, "score": { "total": 0.6749, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.593, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6208, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8267, "spread": 0.14, "weight": 0.2 }, "audience_breadth": { "jev": 0.313, "judge_a": 0.5, "judge_b": 0.48, "mean": 0.4311, "spread": 0.187, "weight": 0.1 }, "practical_relevance": { "jev": 0.545, "judge_a": 0.35, "judge_b": 0.64, "mean": 0.5117, "spread": 0.29, "weight": 0.15 }, "substantiated": { "jev": 0.87, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.79, "spread": 0.3, "weight": 0.1 }, "technical_depth": { "jev": 0.925, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9183, "spread": 0.13, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9367, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.04, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.15, "judge_a": 0.25, "judge_b": 0.08, "mean": 0.16, "spread": 0.17, "weight": -0.15 } } }, "claims": [ { "text": "Snapintosh boots Apple's macOS 27 (build 26A428) on an ASUS Zenbook A16 laptop with a Qualcomm Snapdragon X2 Elite (X2E-96-100).", "kind": "version", "label": "no_source", "drift": null, "article_quote": "Snapintosh boots Apple's macOS 27 (build 26A428) on an ASUS Zenbook A16 laptop with a Qualcomm Snapdragon X2 Elite (X2E-96-100", "source_quote": null, "source_url": null }, { "text": "The Snapintosh monitor runs Apple's stock kernelcache (the J700/T8140 kernel) unmodified by default.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "The monitor runs Apple's stock kernelcache (the J700/T8140 kernel) unmodified by default", "source_quote": null, "source_url": null }, { "text": "Graphics run on the laptop's own Adreno X2-90 GPU through a driver stack written for this project, with no software, CPU or paravirtual rendering path.", "kind": "superlative", "label": "no_source", "drift": null, "article_quote": "There is no software, CPU or paravirtual rendering path", "source_quote": null, "source_url": null }, { "text": "The macOS desktop on the A16 boots with GPU acceleration, and rendering is correct: the pixel checks across the desktop test scenes pass.", "kind": "benchmark", "label": "matches", "drift": null, "article_quote": "Boots to the desktop with GPU acceleration. Rendering is correct: the pixel checks across the desktop test scenes pass", "source_quote": "The desktop renders on the Adreno X2-90 through AdrenoMetal, AdrenoX2 and air2ir3, with no software fallback. A pixel probe over 14 desktop scenes (36 frames) passes with 0 mismatching tiles", "source_url": "https://github.com/Snapintosh/Snapintosh/blob/main/docs/STATUS.md" }, { "text": "The built-in keyboard and touchpad work, including trackpad gestures, Mission Control and app switching.", "kind": "generalization", "label": "matches", "drift": null, "article_quote": "The built-in keyboard and touchpad work, including trackpad gestures, Mission Control and app switching", "source_quote": "Window drags, scrolling, trackpad gestures, Mission Control and Cmd-Tab app switching all work", "source_url": "https://github.com/Snapintosh/Snapintosh/blob/main/docs/STATUS.md" }, { "text": "Wi-Fi scanning, a WPA2 connection, DHCP and ping work through a host-proxy harness, but joining from the macOS Wi-Fi menu does not work yet.", "kind": "generalization", "label": "matches", "drift": null, "article_quote": "Scanning, a WPA2 connection, DHCP and ping work through a host-proxy harness driven from the Mac. Joining from the macOS Wi-Fi menu does not work yet", "source_quote": "Through the host-proxy harness: scan, a WPA2-PSK connection, DHCP and ping", "source_url": "https://github.com/Snapintosh/Snapintosh/blob/main/docs/STATUS.md" }, { "text": "macOS is installed on, and boots from, the internal NVMe storage.", "kind": "generalization", "label": "matches", "drift": null, "article_quote": "macOS is installed on, and boots from, the internal NVMe", "source_quote": "The stock J700/T8140 kernel boots on six vCPUs from macOS installed on the internal NVMe", "source_url": "https://github.com/Snapintosh/Snapintosh/blob/main/docs/STATUS.md" }, { "text": "Audio, cameras, Bluetooth, sleep and external displays are not supported.", "kind": "generalization", "label": "matches", "drift": null, "article_quote": "Audio, cameras, Bluetooth, sleep and external displays are not supported", "source_quote": "Not supported:** audio, cameras, Bluetooth, sleep and wake, external displays, the 120 Hz panel mode, and any other machine or macOS build", "source_url": "https://github.com/Snapintosh/Snapintosh/blob/main/docs/STATUS.md" } ], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207001" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:00.618Z" } }, { "rank": 60, "id": "st_20261024_835", "title": "An LLM judge passed renders with defects I planted", "url": "https://yamanahlawat.in/articles/llm-judge-same-code-different-scores", "author": "Yaman Ahlawat", "site": "Yaman Ahlawat", "published_at": "2026-10-23T16:50:58.000Z", "reading_minutes": 12, "depth": "three_judges", "format": "engineering_writeup", "summary": "A developer's follow-up to an earlier post on an LLM-driven Figma-to-React pipeline reports that the vision-model reviewer's scores vary by about 0.11 when the same render is scored repeatedly, so the 0.85 pass gate and many of the author's earlier comparisons were within noise.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6747, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.52, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.53, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.6, "judge_b": 0.94, "mean": 0.7533, "spread": 0.34, "weight": 0.2 }, "audience_breadth": { "jev": 0.447, "judge_a": 0.3, "judge_b": 0.62, "mean": 0.4556, "spread": 0.32, "weight": 0.1 }, "practical_relevance": { "jev": 0.808, "judge_a": 0.65, "judge_b": 0.88, "mean": 0.7792, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.9122, "spread": 0.137, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.9, "judge_b": 0.88, "mean": 0.9258, "spread": 0.118, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.55, "judge_b": 0.98, "mean": 0.8133, "spread": 0.43, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.34, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.2067, "spread": 0.26, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211931" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:00.628Z" } }, { "rank": 61, "id": "st_20261024_1398", "title": "Why Bother Me Again? Benchmarking Personal Agent Memory for Interaction Preferences", "url": "https://arxiv.org/abs/2610.30485", "author": "Jialin Li, Zhenhao Chen, Hanjun Luo, Hanan Salam", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T22:08:49.000Z", "reading_minutes": 102, "depth": "three_judges", "format": "original_research", "summary": "MemPABench is a new interactive benchmark for testing whether personal AI agents can remember and act on how users want to be assisted, such as when to ask for confirmation, how much to explain, and when to act autonomously.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.674, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.615, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5783, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.63, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.71, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.547, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5556, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.698, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6725, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8867, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.7, "judge_b": 0.91, "mean": 0.86, "spread": 0.27, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8933, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1767, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:00.590Z" } }, { "rank": 62, "id": "st_20261024_1540", "title": "The Delegation Divide: An Open Index of AI-Agent Delegability across Occupations and Countries", "url": "https://arxiv.org/abs/2610.30752", "author": "Cristian Espinal Maya (Universidad Nacional de Colombia), Santiago Jiménez Londoño (Universidad EAFIT)", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T05:22:09.000Z", "reading_minutes": 65, "depth": "three_judges", "format": "original_research", "summary": "The paper introduces the Agent Delegability Index, an open index rating all 18,796 O*NET task statements on how far AI agents can take over each task, then mapping the results to ISCO-08 occupations and 146 countries' employment.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "data" ] }, "score": { "total": 0.6735, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.685, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6717, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.7367, "spread": 0.29, "weight": 0.2 }, "audience_breadth": { "jev": 0.717, "judge_a": 0.6, "judge_b": 0.85, "mean": 0.7222, "spread": 0.25, "weight": 0.1 }, "practical_relevance": { "jev": 0.755, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6783, "spread": 0.28, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.8033, "spread": 0.39, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.6, "judge_b": 0.94, "mean": 0.8458, "spread": 0.398, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9033, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.0767, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.49, "judge_a": 0.45, "judge_b": 0.25, "mean": 0.3967, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:00.600Z" } }, { "rank": 63, "id": "st_20261024_574", "title": "RubyLLM: What's New in 2.1", "url": "https://rubyllm.com/whats-new-in-2-1/", "author": "RubyLLM", "site": "RubyLLM", "published_at": "2026-10-23T13:52:19.000Z", "reading_minutes": 12, "depth": "three_judges", "format": "release", "summary": "RubyLLM 2.1 is a release of the Ruby library for LLM APIs, and its headline changes are mostly performance work that needs no code changes: faster streaming, fewer allocations, shared OAuth tokens, and fewer database queries.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6733, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.62, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5967, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.6, "judge_b": 0.92, "mean": 0.7633, "spread": 0.32, "weight": 0.2 }, "audience_breadth": { "jev": 0.367, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.4956, "spread": 0.253, "weight": 0.1 }, "practical_relevance": { "jev": 0.965, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8483, "spread": 0.265, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7856, "spread": 0.337, "weight": 0.1 }, "technical_depth": { "jev": 0.978, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.9092, "spread": 0.128, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8767, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.45, "judge_b": 0.12, "mean": 0.2067, "spread": 0.4, "weight": -0.15 }, "overclaim": { "jev": 0.15, "judge_a": 0.4, "judge_b": 0.08, "mean": 0.21, "spread": 0.32, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208996" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:00.720Z" } }, { "rank": 64, "id": "st_20261024_193", "title": "Running the Patched PHP 8.3.36 on Ubuntu 20.04 Servers with Docker", "url": "https://www.vincentschmalbach.com/running-patched-php-8-3-36-ubuntu-20-04-docker/", "author": "Vincent Schmalbach", "site": "Vincent Schmalbach", "published_at": "2026-10-23T06:56:50.000Z", "reading_minutes": 7, "depth": "claim_checked", "format": "engineering_writeup", "summary": "Hrih News: A developer documents running the patched PHP 8.3.36 on two production Ubuntu 20.04 Laravel servers by running PHP-FPM from Docker's official image while keeping the host's nginx and other services.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "security", "devops", "it" ] }, "score": { "total": 0.6729, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.495, "judge_a": 0.45, "judge_b": 0.7, "mean": 0.5483, "spread": 0.25, "weight": 0.25 }, "reader_interest": { "jev": 0.57, "judge_a": 0.7, "judge_b": 0.85, "mean": 0.7067, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.423, "judge_a": 0.4, "judge_b": 0.42, "mean": 0.4144, "spread": 0.023, "weight": 0.1 }, "practical_relevance": { "jev": 0.923, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8675, "spread": 0.122, "weight": 0.15 }, "substantiated": { "jev": 0.837, "judge_a": 0.72, "judge_b": 0.78, "mean": 0.7789, "spread": 0.117, "weight": 0.1 }, "technical_depth": { "jev": 1, "judge_a": 0.9, "judge_b": 0.9, "mean": 0.9333, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.8567, "spread": 0.25, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.12, "judge_b": 0.12, "mean": 0.09, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.14, "judge_a": 0.15, "judge_b": 0.12, "mean": 0.1367, "spread": 0.03, "weight": -0.15 } } }, "claims": [ { "text": "PHP fixed CVE-2026-109618 and CVE-2026-109619 on October 21, 2026, in PHP 8.3.36 and the matching 8.2.35, 8.4.27 and 8.5.12 releases.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "PHP fixed the two critical vulnerabilities CVE-2026-109618 and CVE-2026-109619 on October 21, 2026, in PHP 8.3.36 and the matching 8.2.35, 8.4.27 and 8.5.12 releases", "source_quote": null, "source_url": "https://github.com/php/php-src/security-advisories/GHSA-w8hw-w9fx-j79w" }, { "text": "Both vulnerabilities allow remote code execution through requests to the site.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "Both allow remote code execution, which means an attacker can run their own code on the server through requests to your site", "source_quote": null, "source_url": "https://github.com/php/php-src/security-advisories/GHSA-w8hw-w9fx-j79w" }, { "text": "On October 21 the ondrej/php PPA published PHP 8.3.36 for Ubuntu 24.04 and 22.04 but nothing for 20.04; the last PHP 8.3 build for 20.04 was 8.3.21 in May 2025, listed as deleted.", "kind": "date", "label": "couldnt_check", "drift": null, "article_quote": "On October 21 the PPA published 8.3.36 for Ubuntu 24.04 and 22.04. For 20.04 it has nothing. The last PHP 8.3 build for 20.04 was 8.3.21 in May 2025", "source_quote": null, "source_url": "https://launchpad.net/~ondrej/+archive/ubuntu/php" }, { "text": "Ubuntu ended standard security maintenance for 20.04 in May 2025.", "kind": "date", "label": "matches", "drift": null, "article_quote": "Ubuntu ended standard security maintenance for 20.04 in May 2025", "source_quote": "20.04 LTS Ubuntu 20.04 LTS Released: Apr 2020 | Standard security maintenance May 2025 With Ubuntu Pro", "source_url": "https://ubuntu.com/about/release-cycle" }, { "text": "Ubuntu Pro keeps patching Ubuntu's own 20.04 packages until May 2030.", "kind": "date", "label": "no_source", "drift": null, "article_quote": "Ubuntu Pro keeps patching Ubuntu's own 20.04 packages until May 2030", "source_quote": null, "source_url": null }, { "text": "Ubuntu's own 20.04 archive only ships PHP 7.4.", "kind": "version", "label": "no_source", "drift": null, "article_quote": "Ubuntu's own 20.04 archive only ships PHP 7.4", "source_quote": null, "source_url": null }, { "text": "The official image's zz-docker.conf makes PHP-FPM listen on port 9000 on all interfaces.", "kind": "version", "label": "matches", "drift": null, "article_quote": "The image also ships a zz-docker.conf that makes PHP-FPM listen on port 9000 on all interfaces, as the official Dockerfile shows", "source_quote": "echo '; default listen address for easy override in later php-fpm.d/*.conf files'; \\ echo 'listen = 9000'; \\ } | tee php-fpm.d/docker.conf", "source_url": "https://github.com/docker-library/php/blob/master/8.3/bookworm/fpm/Dockerfile" }, { "text": "Running PHP-FPM in the container added about 5 to 10 ms per page; three pages took 109, 102 and 98 ms on the host socket and 113, 111 and 111 ms through the container.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "The container cost me about 5 to 10 ms per page", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50205334" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:00.727Z" } }, { "rank": 65, "id": "st_20261024_1409", "title": "Speculative Acceptance Does Not Decay with Position: A Full-Length Diagnostic on Long Chain-of-Thought with Independent Drafts", "url": "https://arxiv.org/abs/2610.30644", "author": "Zhiming Meng", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T03:05:29.000Z", "reading_minutes": 27, "depth": "three_judges", "format": "original_research", "summary": "A new paper measures whether speculative-decoding acceptance rate falls as a reasoning model's chain-of-thought grows longer, using independent draft models that were never trained on target outputs, a setup prior work lacked.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6729, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.618, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6158, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7133, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.47, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.53, "spread": 0.32, "weight": 0.1 }, "practical_relevance": { "jev": 0.69, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6733, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.8722, "spread": 0.177, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9425, "spread": 0.148, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8633, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.27, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:00.999Z" } }, { "rank": 66, "id": "st_20261024_405", "title": "Arc Decide RAG and conversation insights under 1gb", "url": "https://huggingface.co/Arc53/arc-decide-0.4b-v0.2", "author": null, "site": "huggingface.co", "published_at": "2026-10-23T10:34:36.000Z", "reading_minutes": 11, "depth": "three_judges", "format": "release", "summary": "Arc Decide 0.4B v0.2 is a 0.41B-parameter model, distilled from its 0.6B sibling, that makes per-request decisions for RAG and agent apps (relevance, injection, grounding, routing, conversation analytics) in one forward pass without generating text.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6725, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.535, "judge_a": 0.35, "judge_b": 0.68, "mean": 0.5217, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.71, "judge_a": 0.55, "judge_b": 0.88, "mean": 0.7133, "spread": 0.33, "weight": 0.2 }, "audience_breadth": { "jev": 0.453, "judge_a": 0.4, "judge_b": 0.56, "mean": 0.4711, "spread": 0.16, "weight": 0.1 }, "practical_relevance": { "jev": 0.948, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8358, "spread": 0.248, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.91, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9592, "spread": 0.098, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.93, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.11, "spread": 0.15, "weight": -0.15 }, "overclaim": { "jev": 0.3, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.2433, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206861" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:00.755Z" } }, { "rank": 67, "id": "st_20261024_1556", "title": "OpenAttention: One Percent Is Enough with Tail Proxies", "url": "https://arxiv.org/abs/2610.30951", "author": "Hyunwoo Yu, Jongbeom Lee, Hyunbin Kim et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:03:08.000Z", "reading_minutes": 57, "depth": "three_judges", "format": "original_research", "summary": "OpenAttention proposes approximating the unread tail of the key-value cache in sparse attention with a few proxy tokens, rather than discarding or randomly sampling it.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6723, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.708, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6458, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.6933, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.527, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5156, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.783, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7342, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.8633, "spread": 0.23, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9567, "spread": 0.09, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.89, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.0633, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.37, "judge_a": 0.35, "judge_b": 0.3, "mean": 0.34, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.091Z" } }, { "rank": 68, "id": "st_20261024_1513", "title": "Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems", "url": "https://arxiv.org/abs/2610.30418", "author": "Daniel Ruiz, Luca Luceri, Emilio Ferrara", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T20:25:11.000Z", "reading_minutes": 27, "depth": "three_judges", "format": "original_research", "summary": "The paper \"Agent Mosaic\" studies whether LLM agents leak confidential conclusions even when every source document they read is individually accessible to the user. Using 30 synthetic cybersecurity, medical, and military examples, the authors measure how often agents disclose a designated higher-classified composite under benign, explicit, and adversarial queries, in…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6721, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.688, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6358, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.7233, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.707, "judge_a": 0.5, "judge_b": 0.56, "mean": 0.5889, "spread": 0.207, "weight": 0.1 }, "practical_relevance": { "jev": 0.715, "judge_a": 0.6, "judge_b": 0.68, "mean": 0.665, "spread": 0.115, "weight": 0.15 }, "substantiated": { "jev": 0.95, "judge_a": 0.7, "judge_b": 0.79, "mean": 0.8133, "spread": 0.25, "weight": 0.1 }, "technical_depth": { "jev": 0.965, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.8083, "spread": 0.365, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8567, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2167, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.107Z" } }, { "rank": 69, "id": "st_20261024_852", "title": "Beyond g-mean²: complementary metrics for monitorability", "url": "https://alignment.openai.com/beyond-gmean-squared/", "author": null, "site": "alignment.openai.com", "published_at": "2026-10-23T16:35:59.000Z", "reading_minutes": 23, "depth": "three_judges", "format": "original_research", "summary": "This post extends the g-mean² monitorability metric with two complementary measures. Pair-matched comparison evaluates models only on task instances both can validly exhibit the target behavior, so differences in which tasks each model fails on don't drive the ranking.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6716, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.6, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6067, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.57, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.6833, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.507, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.4956, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.67, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.6567, "spread": 0.1, "weight": 0.15 }, "substantiated": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.8, "mean": 0.8433, "spread": 0.13, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9592, "spread": 0.098, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.1, "judge_b": 0.04, "mean": 0.06, "spread": 0.06, "weight": -0.15 }, "overclaim": { "jev": 0.2, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1733, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211679" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:00.890Z" } }, { "rank": 70, "id": "st_20261024_1571", "title": "Prompting Strategy Effects Depend on the Judge: A Multilingual Study of LLM-as-a-Judge Robustness", "url": "https://arxiv.org/abs/2610.31172", "author": "Asal Mehrabi, Mohammad Aliannejadi, Clemencia Siro", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T12:27:05.000Z", "reading_minutes": 38, "depth": "three_judges", "format": "original_research", "summary": "This study benchmarks nine prompting strategies for LLM-as-a-judge evaluation across four open-weight judge models, 27 languages, four tasks and 13 datasets, producing over one million pointwise scores on gold-reference outputs validated against human judgments.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6715, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.703, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.6075, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.6933, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.587, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6022, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.75, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.71, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.8, "judge_b": 0.8, "mean": 0.8644, "spread": 0.193, "weight": 0.1 }, "technical_depth": { "jev": 0.98, "judge_a": 0.75, "judge_b": 0.92, "mean": 0.8833, "spread": 0.23, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.89, "spread": 0.15, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.3, "judge_b": 0.34, "mean": 0.2967, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.222Z" } }, { "rank": 71, "id": "st_20261024_1597", "title": "Predictable performance is suboptimal, and what it means for model routing", "url": "https://arxiv.org/abs/2610.31428", "author": "Guanhua Zhang, Moritz Hardt", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:27:27.000Z", "reading_minutes": 56, "depth": "three_judges", "format": "original_research", "summary": "Researchers at the Max Planck Institute for Intelligent Systems argue that a language model's errors are predictable beyond its own confidence mainly when the model is suboptimal, so routing systems that use separate performance predictors gain most on weaker models.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.671, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.633, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6008, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.6933, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.59, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.5233, "spread": 0.19, "weight": 0.1 }, "practical_relevance": { "jev": 0.655, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6417, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8833, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.968, "judge_a": 0.85, "judge_b": 0.96, "mean": 0.9258, "spread": 0.118, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8567, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0267, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.3, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1933, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.281Z" } }, { "rank": 72, "id": "st_20261024_1493", "title": "Self-Evolving Static Analysis via LLM Pseudo-Labeling", "url": "https://arxiv.org/abs/2610.30184", "author": "Rui Melo, Andre Duarte, Aditya Oke et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-20T10:17:26.000Z", "reading_minutes": 60, "depth": "three_judges", "format": "original_research", "summary": "SESAA is a framework in which an LLM agent starts with no static-analysis rules and iteratively evolves readable Joern queries, using only pseudo-labels from a second LLM critic and no ground-truth labels.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai", "security" ] }, "score": { "total": 0.6698, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.6, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6433, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.7667, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.613, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.5544, "spread": 0.113, "weight": 0.1 }, "practical_relevance": { "jev": 0.605, "judge_a": 0.65, "judge_b": 0.75, "mean": 0.6683, "spread": 0.145, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.8022, "spread": 0.347, "weight": 0.1 }, "technical_depth": { "jev": 0.985, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.915, "spread": 0.185, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9267, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0433, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.51, "judge_a": 0.45, "judge_b": 0.2, "mean": 0.3867, "spread": 0.31, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.404Z" } }, { "rank": 73, "id": "st_20261024_484", "title": "Disgraced eMudhra drops out of TLS business after using cert private key of 4", "url": "https://arstechnica.com/security/2026/10/disgraced-emudhra-drops-out-of-tls-business-following-certificate-debacle/", "author": null, "site": "Ars Technica", "published_at": "2026-10-23T11:01:56.000Z", "reading_minutes": 1, "depth": "three_judges", "format": "news_report", "summary": "eMudhra, the India-based certificate authority that issued eight TLS intermediate certificates with private keys so weak they could be cracked in milliseconds, says it is exiting the global SSL/TLS business.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security" ] }, "score": { "total": 0.6693, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.638, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6225, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.8333, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.653, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6911, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.563, "judge_a": 0.55, "judge_b": 0.8, "mean": 0.6375, "spread": 0.25, "weight": 0.15 }, "substantiated": { "jev": 0.53, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.69, "spread": 0.31, "weight": 0.1 }, "technical_depth": { "jev": 0.728, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6825, "spread": 0.128, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.94, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.13, "judge_a": 0.03, "judge_b": 0.02, "mean": 0.06, "spread": 0.11, "weight": -0.15 }, "overclaim": { "jev": 0.48, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.2667, "spread": 0.36, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207082" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:01.234Z" } }, { "rank": 74, "id": "st_20261024_1064", "title": "PHP advisory: disable session upload progress to prevent another unlikely RCE", "url": "https://www.php.net/archive/2026.php", "author": null, "site": "php.net", "published_at": "2026-10-23T17:58:38.000Z", "reading_minutes": 27, "depth": "three_judges", "format": "news_report", "summary": "PHP's maintainers published a security advisory and new releases (8.5.12, 8.4.27, 8.3.36, 8.2.35) fixing two critical flaws, CVE-2026-109618 and CVE-2026-109619, that could allow unauthenticated remote code execution through out-of-memory corruption and stale global state.", "also_covered_by": [ { "site": "Calif", "url": "https://calif.io/php2shell.x", "title": "A tale of two PHP remote vulnerabilities" }, { "site": "jacobg.co", "url": "https://jacobg.co/ai-agents/", "title": "AI Agents Kinda Broke the Internet" }, { "site": "GitHub", "url": "https://github.com/wasdthedebugger/php2shell", "title": "PHP2SHELL Patch for Older Versions" }, { "site": "SazM", "url": "https://sazm.in/articles/php2shell-cve-2026-109618-mitigation", "title": "PHP2Shell (CVE-2026-109618): Zero-Downtime Mitigations and Upgrade Runbook" }, { "site": "Drew's Dev Blog", "url": "https://drew.silcock.dev/blog/is-this-the-new-cyber-normal/", "title": "Is this the new cyber normal?" }, { "site": "Vu Agency", "url": "https://vu.co.uk/blog/who-has-patched-php", "title": "Who has patched PHP2Shell? I checked nearly 1M websites" } ], "labels": { "content_type": "news_report", "project_stage": "product_release", "categories": [ "programming", "security" ] }, "score": { "total": 0.6693, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.53, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6367, "spread": 0.25, "weight": 0.25 }, "reader_interest": { "jev": 0.53, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.71, "spread": 0.37, "weight": 0.2 }, "audience_breadth": { "jev": 0.563, "judge_a": 0.67, "judge_b": 0.48, "mean": 0.5711, "spread": 0.19, "weight": 0.1 }, "practical_relevance": { "jev": 0.86, "judge_a": 0.85, "judge_b": 0.88, "mean": 0.8633, "spread": 0.03, "weight": 0.15 }, "substantiated": { "jev": 0.737, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.7789, "spread": 0.2, "weight": 0.1 }, "technical_depth": { "jev": 0.84, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.7033, "spread": 0.29, "weight": 0.1 }, "new_information": { "jev": 0.85, "judge_a": 0.6, "judge_b": 0.99, "mean": 0.8133, "spread": 0.39, "weight": 0.1 }, "promotional": { "jev": 0.12, "judge_a": 0.1, "judge_b": 0.02, "mean": 0.08, "spread": 0.1, "weight": -0.15 }, "overclaim": { "jev": 0.44, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.24, "spread": 0.36, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": true, "url": "https://news.ycombinator.com/item?id=50212942" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, front page", "checked_at": "2026-10-24T08:54:01.349Z" } }, { "rank": 75, "id": "st_20261024_1482", "title": "Who Is Sycophantic to Whom? Confident LLMs, Human Deference, and the Collapse of Trust Calibration", "url": "https://arxiv.org/abs/2610.31579", "author": "Deepthika S, Pawni Yadav, B. Sankar", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:15:59.000Z", "reading_minutes": 80, "depth": "three_judges", "format": "original_research", "summary": "A new arXiv paper by Deepthika S and B. Sankar tests whether users defer to confident, consistent LLMs even when those models are wrong. In a within-subjects study of 32 participants learning science topics from assistants with 0 to 100 percent planted falsehoods, quiz scores fell from 7.4 to 2.9…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.669, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.698, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6425, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.93, "mean": 0.73, "spread": 0.33, "weight": 0.2 }, "audience_breadth": { "jev": 0.773, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6811, "spread": 0.223, "weight": 0.1 }, "practical_relevance": { "jev": 0.793, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7242, "spread": 0.193, "weight": 0.15 }, "substantiated": { "jev": 0.96, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.8333, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.975, "judge_a": 0.35, "judge_b": 0.89, "mean": 0.7383, "spread": 0.625, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.86, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.41, "judge_a": 0.3, "judge_b": 0.32, "mean": 0.3433, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.622Z" } }, { "rank": 76, "id": "st_20261024_1272", "title": "AI Just Raised the Stakes on Continuous Delivery", "url": "https://www.bradoyler.com/ai/continuous-delivery-is-a-security-feature/", "author": null, "site": "{ Brad Oyler }", "published_at": "2026-10-23T21:53:52.000Z", "reading_minutes": 10, "depth": "three_judges", "format": "analysis_opinion", "summary": "A week of critical flaws across PHP, Node.js, Plex, an Indian certificate authority, and the LiteLLM AI proxy highlighted how quickly AI-assisted discovery and exploitation are moving.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai", "security", "devops" ] }, "score": { "total": 0.6687, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.718, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6292, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.83, "judge_a": 0.72, "judge_b": 0.84, "mean": 0.7967, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.987, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8222, "spread": 0.287, "weight": 0.1 }, "practical_relevance": { "jev": 0.898, "judge_a": 0.72, "judge_b": 0.82, "mean": 0.8125, "spread": 0.178, "weight": 0.15 }, "substantiated": { "jev": 0.66, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.63, "spread": 0.13, "weight": 0.1 }, "technical_depth": { "jev": 0.79, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.62, "spread": 0.34, "weight": 0.1 }, "new_information": { "jev": 0.85, "judge_a": 0.5, "judge_b": 0.96, "mean": 0.77, "spread": 0.46, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.05, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.35, "judge_b": 0.31, "mean": 0.31, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50215890" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:01.493Z" } }, { "rank": 77, "id": "st_20261024_1223", "title": "From 27 minutes to 4: Cutting CI build time by 85%", "url": "https://alexitc.com/blog/2026-10-23-from-27-minutes-to-4-cutting-ci-build-time-by-85-percent/", "author": null, "site": "alexitc.com", "published_at": "2026-10-23T19:08:00.000Z", "first_published_at": "2026-10-23T00:00:00.000Z", "reading_minutes": 7, "depth": "claim_checked", "format": "engineering_writeup", "summary": "A developer documents cutting the CI build time of their open-source Scala expense-tracker from 27 minutes to about 4 minutes, roughly 85%. The biggest wins came from sharing one Postgres test container across the run, booting one app per test suite instead of per test, skipping work for unaffected code…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "devops" ] }, "score": { "total": 0.6675, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.505, "judge_a": 0.3, "judge_b": 0.48, "mean": 0.4283, "spread": 0.205, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7533, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.78, "judge_a": 0.6, "judge_b": 0.63, "mean": 0.67, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.985, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.895, "spread": 0.185, "weight": 0.15 }, "substantiated": { "jev": 0.96, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8467, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.975, "judge_a": 0.75, "judge_b": 0.78, "mean": 0.835, "spread": 0.225, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.5, "judge_b": 0.95, "mean": 0.7833, "spread": 0.45, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1133, "spread": 0.14, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.2, "judge_b": 0.06, "mean": 0.14, "spread": 0.14, "weight": -0.15 } } }, "claims": [ { "text": "The expense-tracker project's CI execution time grew to 27 minutes per run and was reduced to an average of 4 minutes.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "the CI execution time grew to 27 minutes per run! Of course this is an annoying bottleneck. With some effort I managed to get this down to an average of…", "source_quote": null, "source_url": null }, { "text": "Switching from one Postgres container per test to one shared container for the whole run saved about 15 minutes.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "I switched from launching a container per test to one shared container for the whole run (amazing result, 15 minutes saved", "source_quote": null, "source_url": "https://github.com/AlexITC/expense-tracker/pull/22" }, { "text": "Booting one Play app per suite instead of per test cut about 70 seconds from the server test step (183s to 113s).", "kind": "number", "label": "matches", "drift": null, "article_quote": "Booting one app per suite instead of per test was trivial and cut ~70 seconds", "source_quote": "Before (`GuiceOneServerPerTest`) | ~183s | | After (`GuiceOneServerPerSuite`) | ~113s", "source_url": "https://github.com/AlexITC/expense-tracker/pull/27" }, { "text": "Running the tests in parallel saved about 43 seconds on the server test step (122s to 79s).", "kind": "number", "label": "matches", "drift": null, "article_quote": "Running the test suite in parallel saved another 40 seconds", "source_quote": "35% faster on the 4-core runner (test step 122s→79s", "source_url": "https://github.com/AlexITC/expense-tracker/pull/37" }, { "text": "Using real bcrypt at the cheapest cost setting in tests saved about 25 seconds on the HTTP-level test suite (59s to 34s).", "kind": "number", "label": "matches", "drift": null, "article_quote": "I kept real bcrypt in the tests but used the weakest cost setting, saving ~25 seconds without faking the hashing", "source_quote": "10 (before) | 59s | | 4 (after) | 33–34s | **~25s faster** on the Play specs", "source_url": "https://github.com/AlexITC/expense-tracker/pull/38" }, { "text": "A cache step in the CI workflow ran but never persisted its cache because every save failed, so runs repeated work that should have been skipped.", "kind": "causal", "label": "couldnt_check", "drift": null, "article_quote": "So the cache was never persisted; every run re-did the work it was supposed to skip", "source_quote": null, "source_url": "https://github.com/AlexITC/expense-tracker/pull/33" }, { "text": "On GitHub Actions, a cache saved on a PR branch is only visible to that PR, not to main or other PRs.", "kind": "generalization", "label": "matches", "drift": null, "article_quote": "on GitHub Actions a cache saved on a PR branch is only visible to that PR, not to main or other PRs", "source_quote": "When a cache is created by a workflow run triggered on a pull request, the cache is created for the merge ref (refs/pull/.../merge). Because of this, the cache will have…", "source_url": "https://docs.github.com/en/actions/reference/workflows-and-actions/dependency-caching" }, { "text": "The end-to-end CI improvement was from about 27 minutes to about 4 minutes, roughly 85%.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "the end-to-end result is ~27 min -> ~4 min (~85%", "source_quote": null, "source_url": null } ], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213901" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:01.508Z" } }, { "rank": 78, "id": "st_20261024_1416", "title": "Faithful Latent Harm Monitors Work No Better Than the Models Own Behaviour at a Matched False Alarm Rate", "url": "https://arxiv.org/abs/2610.30727", "author": "Abhishek Mishra, Sagnik Chatterjee, Suyash Maniyar", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T04:54:53.000Z", "reading_minutes": 56, "depth": "three_judges", "format": "original_research", "summary": "This paper evaluates \"latent harm monitors,\" which read harmfulness from LLM activations to flag harmful prompts even when the model complies. Across 15 models and 38 activation-based readers, the authors compare these monitors against TF-IDF and sentence-embedding classifiers, an independent safety classifier, and each model's own refusal, using reworded and…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6673, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.66, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.5967, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6867, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.593, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.5244, "spread": 0.193, "weight": 0.1 }, "practical_relevance": { "jev": 0.673, "judge_a": 0.5, "judge_b": 0.65, "mean": 0.6075, "spread": 0.173, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.85, "judge_b": 0.88, "mean": 0.9078, "spread": 0.143, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.9275, "spread": 0.143, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8567, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.02, "judge_b": 0.01, "mean": 0.0233, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.19, "spread": 0.21, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.770Z" } }, { "rank": 79, "id": "st_20261024_417", "title": "How Generative AI Is Scaling Transnational Repression", "url": "https://www.justsecurity.org/159245/how-generative-ai-scaling-transnational-repression/", "author": "Kate Pundyk, Siena Anstis, Marcus Michaelsen", "site": "Just Security", "published_at": "2026-10-23T10:15:11.000Z", "reading_minutes": 9, "depth": "three_judges", "format": "analysis_opinion", "summary": "Anthropic's September 2026 Threat Intelligence report documents how a China-aligned operator used its Claude model to help track Uyghurs living in Syria, translating messages, geolocating targets, and identifying vulnerable people as potential informants.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "ai", "security", "policy" ] }, "score": { "total": 0.6671, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.738, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7192, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.75, "judge_b": 0.9, "mean": 0.8067, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.673, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6644, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.753, "judge_a": 0.45, "judge_b": 0.76, "mean": 0.6542, "spread": 0.31, "weight": 0.15 }, "substantiated": { "jev": 0.63, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7033, "spread": 0.15, "weight": 0.1 }, "technical_depth": { "jev": 0.688, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5892, "spread": 0.288, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.97, "mean": 0.8667, "spread": 0.27, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.07, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.35, "judge_b": 0.2, "mean": 0.2933, "spread": 0.15, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 8, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206724" }, "lobsters": null, "line": "Elsewhere: HN 8 pts, not on front page", "checked_at": "2026-10-24T08:54:01.627Z" } }, { "rank": 80, "id": "st_20261024_1620", "title": "SIM-UT: A User Simulation based Evaluation Framework for Medical Triage Chatbots", "url": "https://arxiv.org/abs/2610.30412", "author": "Deepali Mittal, Dheeraj Mundhra, Yulia Sidorova et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T20:14:16.000Z", "reading_minutes": 39, "depth": "three_judges", "format": "original_research", "summary": "SIM-UT is a simulation-based evaluation framework that tests medical triage chatbots through multi-turn, multilingual conversations with realistic and adversarial simulated users, rather than single-turn knowledge benchmarks.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.667, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.658, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6258, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.7533, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.51, "judge_a": 0.5, "judge_b": 0.52, "mean": 0.51, "spread": 0.02, "weight": 0.1 }, "practical_relevance": { "jev": 0.835, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7717, "spread": 0.135, "weight": 0.15 }, "substantiated": { "jev": 0.933, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7844, "spread": 0.333, "weight": 0.1 }, "technical_depth": { "jev": 0.895, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7583, "spread": 0.295, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.8833, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.15, "judge_b": 0.03, "mean": 0.0833, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.3, "judge_b": 0.2, "mean": 0.2467, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.806Z" } }, { "rank": 81, "id": "st_20261024_1621", "title": "Rethinking the Evaluation of LLM Fact-Checkers", "url": "https://arxiv.org/abs/2610.30504", "author": "Ayana Niwa, Miyu Yamada, Kouta Nakayama et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T22:36:44.000Z", "reading_minutes": 40, "depth": "three_judges", "format": "original_research", "summary": "JSFactCheckBench is a Japanese benchmark for fact-checking LLM answers that makes three evaluation choices explicit: claims drawn from answers to real misleading-information questions, factuality judged at a stated reference time with claim-importance labels, and evidence limited to controlled Web sources.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6665, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.635, "judge_a": 0.35, "judge_b": 0.74, "mean": 0.575, "spread": 0.39, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.55, "judge_b": 0.86, "mean": 0.6733, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.563, "judge_a": 0.35, "judge_b": 0.61, "mean": 0.5078, "spread": 0.26, "weight": 0.1 }, "practical_relevance": { "jev": 0.723, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.6942, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.9067, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.75, "judge_b": 0.9, "mean": 0.88, "spread": 0.24, "weight": 0.1 }, "new_information": { "jev": 0.84, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.84, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.15, "judge_a": 0.25, "judge_b": 0.08, "mean": 0.16, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.960Z" } }, { "rank": 82, "id": "st_20261024_1436", "title": "Measuring the Reasoning-Action Coupling in Driving Vision-Language-Action Models through Counterfactuals", "url": "https://arxiv.org/abs/2610.30953", "author": "Gabriel Manalu, Saad Wazir, Thao Do et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:04:06.000Z", "reading_minutes": 30, "depth": "three_judges", "format": "original_research", "summary": "This paper uses counterfactual interventions on four driving vision-language-action models to test whether their generated reasoning actually drives their trajectories and whether both respond to changes in the visual scene. Its tests cover reasoning-side text edits compared against camera blanking, and object removal from camera frames.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6661, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.65, "judge_a": 0.45, "judge_b": 0.82, "mean": 0.64, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.6967, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.49, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.52, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.643, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6208, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.967, "judge_a": 0.72, "judge_b": 0.9, "mean": 0.8622, "spread": 0.247, "weight": 0.1 }, "technical_depth": { "jev": 0.978, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.9125, "spread": 0.178, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8667, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.2467, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:01.921Z" } }, { "rank": 83, "id": "st_20261024_1669", "title": "The Fusion Tax: A Diagnostic Analysis of When Hybrid Retrieval Hurts", "url": "https://arxiv.org/abs/2610.30612", "author": "Aaditya Chauhan, Arun Agarwal", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T02:24:08.000Z", "depth": "three_judges", "format": "original_research", "summary": "A September 2026 study examines the costs of hybrid retrieval, which combines lexical BM25 with dense neural search and is widely recommended on the strength of average gains.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "data" ] }, "score": { "total": 0.666, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.643, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5708, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.72, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.477, "judge_a": 0.4, "judge_b": 0.55, "mean": 0.4756, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.823, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7675, "spread": 0.123, "weight": 0.15 }, "substantiated": { "jev": 0.837, "judge_a": 0.75, "judge_b": 0.78, "mean": 0.7889, "spread": 0.087, "weight": 0.1 }, "technical_depth": { "jev": 0.833, "judge_a": 0.85, "judge_b": 0.85, "mean": 0.8442, "spread": 0.017, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8733, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0333, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.25, "judge_b": 0.12, "mean": 0.1933, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.053Z" } }, { "rank": 84, "id": "st_20261024_1734", "title": "nnInteractive Agent: General-Purpose AI Agents as 3D Image Annotators", "url": "https://arxiv.org/abs/2610.31080", "author": "Fabian Isensee, Constantin Ulrich Harsy, Maximilian Rokuss et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T10:55:15.000Z", "depth": "three_judges", "format": "original_research", "summary": "The paper presents nninteractive-agent, which lets general-purpose AI agents such as Claude Code and Codex perform 3D medical and biological image segmentation from plain-language requests, using the nnInteractive model as tools over the Model Context Protocol and no extra training.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.666, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.6, "judge_a": 0.55, "judge_b": 0.75, "mean": 0.6333, "spread": 0.2, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.6, "judge_b": 0.95, "mean": 0.7567, "spread": 0.35, "weight": 0.2 }, "audience_breadth": { "jev": 0.43, "judge_a": 0.45, "judge_b": 0.55, "mean": 0.4767, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.758, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7258, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.8, "mean": 0.81, "spread": 0.23, "weight": 0.1 }, "technical_depth": { "jev": 0.728, "judge_a": 0.65, "judge_b": 0.82, "mean": 0.7325, "spread": 0.17, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.85, "judge_b": 0.97, "mean": 0.92, "spread": 0.12, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.1, "judge_b": 0.05, "mean": 0.08, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.23, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.133Z" } }, { "rank": 85, "id": "st_20261024_1495", "title": "Why Constrained Updates Still Forget: A Functional-Path Analysis of Continual SFT", "url": "https://arxiv.org/abs/2610.30191", "author": "Qingsen Ma, Dianyun Wang, Jiahong Liu, Irwin King", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T03:57:12.000Z", "reading_minutes": 88, "depth": "three_judges", "format": "original_research", "summary": "This paper analyzes why constrained updates still cause forgetting during supervised fine-tuning of language models, decomposing old-behavior loss changes into constraint, probe mismatch, gradient staleness, and finite-step terms.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6648, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.65, "judge_a": 0.35, "judge_b": 0.78, "mean": 0.5933, "spread": 0.43, "weight": 0.25 }, "reader_interest": { "jev": 0.53, "judge_a": 0.55, "judge_b": 0.9, "mean": 0.66, "spread": 0.37, "weight": 0.2 }, "audience_breadth": { "jev": 0.583, "judge_a": 0.3, "judge_b": 0.72, "mean": 0.5344, "spread": 0.42, "weight": 0.1 }, "practical_relevance": { "jev": 0.643, "judge_a": 0.5, "judge_b": 0.8, "mean": 0.6475, "spread": 0.3, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.87, "spread": 0.17, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9575, "spread": 0.093, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8867, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0267, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2233, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.240Z" } }, { "rank": 86, "id": "st_20261024_954", "title": "W3C: A request isn’t a blank check for AI agents", "url": "https://www.heise.de/en/news/W3C-A-request-isn-t-a-blank-check-for-AI-agents-11498111.html", "author": "heise online", "site": "heise online English", "published_at": "2026-10-23T12:18:00.000Z", "first_published_at": "2026-10-23T12:18:00.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "news_report", "summary": "The W3C strategy team has published a discussion paper proposing that web standards should constrain AI agents that now act on users' behalf, not just read content.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai", "security", "policy" ] }, "score": { "total": 0.6637, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.683, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6175, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7633, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.823, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7344, "spread": 0.223, "weight": 0.1 }, "practical_relevance": { "jev": 0.695, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.655, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.51, "judge_a": 0.5, "judge_b": 0.8, "mean": 0.6033, "spread": 0.3, "weight": 0.1 }, "technical_depth": { "jev": 0.675, "judge_a": 0.6, "judge_b": 0.68, "mean": 0.6517, "spread": 0.08, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.87, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0467, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.1367, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.273Z" } }, { "rank": 87, "id": "st_20261024_1590", "title": "HERD: Harness-Evolving Recursive Distillation for Continual Learning", "url": "https://arxiv.org/abs/2610.31401", "author": "Xun Sun, Thanh Thi Nguyen, Thuy-Trang Vu", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:03:20.000Z", "reading_minutes": 60, "depth": "three_judges", "format": "original_research", "summary": "HERD is a hybrid continual learning method that pairs an evolving harness of reusable skills and a bounded memory of hard examples with recursive distillation into model weights, under drift constraints anchored to the previous checkpoint and the base model.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6636, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.65, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6433, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.6, "judge_b": 0.94, "mean": 0.7133, "spread": 0.34, "weight": 0.2 }, "audience_breadth": { "jev": 0.537, "judge_a": 0.4, "judge_b": 0.73, "mean": 0.5556, "spread": 0.33, "weight": 0.1 }, "practical_relevance": { "jev": 0.555, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.575, "spread": 0.27, "weight": 0.15 }, "substantiated": { "jev": 0.95, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.79, "spread": 0.35, "weight": 0.1 }, "technical_depth": { "jev": 0.978, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9192, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9133, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2533, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.349Z" } }, { "rank": 88, "id": "st_20261024_1561", "title": "Mostly Reading: What Open-Book Scores Measure in Document-Level LoRA Adapters", "url": "https://arxiv.org/abs/2610.31017", "author": "Martino M. L. Pulici, Cuong Xuan Chu, Evgeny Kharlamov, Volker Tresp", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:52:29.000Z", "reading_minutes": 75, "depth": "three_judges", "format": "original_research", "summary": "Researchers testing single-document LoRA adapters on Llama-3.1-8B-Instruct found that open-book scores mostly reflect a shared reading ability rather than stored document content. A random adapter recovered 82% of a specialist's open-book advantage but only 2% of its cloze advantage, which probes recall without the source text.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6635, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.653, "judge_a": 0.35, "judge_b": 0.82, "mean": 0.6075, "spread": 0.47, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.4, "judge_b": 0.94, "mean": 0.6467, "spread": 0.54, "weight": 0.2 }, "audience_breadth": { "jev": 0.48, "judge_a": 0.3, "judge_b": 0.67, "mean": 0.4833, "spread": 0.37, "weight": 0.1 }, "practical_relevance": { "jev": 0.678, "judge_a": 0.45, "judge_b": 0.82, "mean": 0.6492, "spread": 0.37, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.9022, "spread": 0.177, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9417, "spread": 0.145, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8667, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.2, "judge_b": 0.15, "mean": 0.1967, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.426Z" } }, { "rank": 89, "id": "st_20261024_721", "title": "Apple ordered Allow alternative App Stores - Violated Australian Competition Law", "url": "https://www.abc.net.au/news/2026-10-23/fortnite-epic-games-back-in-apple-app-store/107302160", "author": "national technology reporter Ange Lavoipierre", "site": "abc.net.au", "published_at": "2026-10-23T14:05:18.000Z", "first_published_at": "2026-10-23T00:00:00.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "news_report", "summary": "The Federal Court has found Apple breached Australian competition law by restricting app distribution and payments on iPhones, and ordered it to allow alternative app stores and payment methods by March 2027. The ruling affects an estimated 15 million local users and follows a case brought by Epic Games.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "policy" ] }, "score": { "total": 0.6632, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.96, "judge_a": 0.62, "judge_b": 0.82, "mean": 0.8, "spread": 0.34, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8067, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.507, "judge_a": 0.67, "judge_b": 0.65, "mean": 0.6089, "spread": 0.163, "weight": 0.1 }, "practical_relevance": { "jev": 0.72, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.6733, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.463, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.5978, "spread": 0.317, "weight": 0.1 }, "technical_depth": { "jev": 0.475, "judge_a": 0.2, "judge_b": 0.22, "mean": 0.2983, "spread": 0.275, "weight": 0.1 }, "new_information": { "jev": 0.95, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9133, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.03, "mean": 0.0367, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.2, "judge_b": 0.2, "mean": 0.2367, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209222" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, not on front page", "checked_at": "2026-10-24T08:54:02.364Z" } }, { "rank": 90, "id": "st_20261024_1501", "title": "Automated screening of privacy risks in medical AI", "url": "https://arxiv.org/abs/2610.30229", "author": "Konstantinos Vrettos, Eleftherios Tzanis, Haridimos Kondylakis et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T17:52:43.000Z", "depth": "three_judges", "format": "original_research", "summary": "Researchers measured how vulnerable individual patients are to membership inference attacks, which can reveal whether a patient's records were used to train a medical AI model, across 7,789 models on six clinical datasets.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6632, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.63, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6533, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.7033, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.463, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5278, "spread": 0.157, "weight": 0.1 }, "practical_relevance": { "jev": 0.728, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7025, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8033, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.668, "judge_a": 0.7, "judge_b": 0.7, "mean": 0.6892, "spread": 0.032, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9033, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0467, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.2, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.21, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.596Z" } }, { "rank": 91, "id": "st_20261024_1479", "title": "Beyond Causal Masks: Auditing and Localizing Hidden Future-Token Leakage in Adaptive Sequence Models", "url": "https://arxiv.org/abs/2610.31522", "author": "Huiting Liu, Jinsong Wu", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T16:34:32.000Z", "reading_minutes": 40, "depth": "three_judges", "format": "original_research", "summary": "A paper in Neurocomputing examines hidden future-token leakage, where an autoregressive model reads information from later tokens through a non-causal auxiliary path even when its attention mask is correctly causal, producing misleadingly strong teacher-forced results.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6629, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.62, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5667, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.58, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.6867, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.547, "judge_a": 0.5, "judge_b": 0.63, "mean": 0.5589, "spread": 0.13, "weight": 0.1 }, "practical_relevance": { "jev": 0.738, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7392, "spread": 0.08, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.7, "judge_b": 0.76, "mean": 0.7989, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.8, "judge_b": 0.92, "mean": 0.9042, "spread": 0.193, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.97, "mean": 0.8533, "spread": 0.27, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2167, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.659Z" } }, { "rank": 92, "id": "st_20261024_490", "title": "PHP 8.5 reimplemented from scratch in Rust", "url": "https://github.com/francescotinti/php-rust", "author": null, "site": "GitHub", "published_at": "2026-10-23T12:57:19.000Z", "reading_minutes": 11, "depth": "three_judges", "format": "engineering_writeup", "summary": "phpr is a PHP 8.5 runtime reimplemented from scratch in Rust, judged by the official .phpt test corpus and real-world output rather than Zend's internals. Its bytecode VM reportedly runs the full WordPress core test suite at effective parity against the reference interpreter, with Composer, PHPUnit and Doctrine also passing.…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "early_project", "categories": [ "programming" ] }, "score": { "total": 0.6623, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.708, "judge_a": 0.62, "judge_b": 0.78, "mean": 0.7025, "spread": 0.16, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.8433, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.587, "judge_a": 0.67, "judge_b": 0.72, "mean": 0.6589, "spread": 0.133, "weight": 0.1 }, "practical_relevance": { "jev": 0.593, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.5742, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.72, "judge_b": 0.74, "mean": 0.7989, "spread": 0.217, "weight": 0.1 }, "technical_depth": { "jev": 0.968, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9158, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 1, "mean": 0.94, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.1533, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.46, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.51, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208271" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:02.494Z" } }, { "rank": 93, "id": "st_20261024_527", "title": "AI Identity Management System", "url": "https://datatracker.ietf.org/doc/draft-ietf-wimse-aims/00/", "author": "Aaron Parecki", "site": "IETF Datatracker", "published_at": "2026-10-23T12:28:11.000Z", "reading_minutes": 39, "depth": "three_judges", "format": "other", "summary": "The draft \"AI Identity Management System\" (draft-ietf-wimse-aims-00), an IETF Internet-Draft from the WIMSE working group, proposes best practices for authenticating and authorizing AI agents by composing existing standards such as WIMSE and the OAuth 2.0 family rather than defining new protocols.", "also_covered_by": [], "labels": { "content_type": "other", "project_stage": "not_a_project", "categories": [ "ai", "security", "devops" ] }, "score": { "total": 0.6622, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.625, "judge_a": 0.3, "judge_b": 0.78, "mean": 0.5683, "spread": 0.48, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.6867, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.67, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6667, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.795, "judge_a": 0.6, "judge_b": 0.84, "mean": 0.745, "spread": 0.24, "weight": 0.15 }, "substantiated": { "jev": 0.5, "judge_a": 0.45, "judge_b": 0.87, "mean": 0.6067, "spread": 0.42, "weight": 0.1 }, "technical_depth": { "jev": 0.88, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8533, "spread": 0.08, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8633, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1467, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207901" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:02.564Z" } }, { "rank": 94, "id": "st_20261024_1406", "title": "Model Organisms of Collusion: Corrective Transfer in Untrusted Monitors", "url": "https://arxiv.org/abs/2610.30582", "author": "Juan-Pablo Rivera, Ari Spiesberger, Juan J Vazquez et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T01:22:48.000Z", "reading_minutes": 68, "depth": "three_judges", "format": "original_research", "summary": "Researchers at Arb Research, Redwood Research, Meridian Cambridge, Georgia Tech and Anthropic study whether retraining an AI monitor on one caught collusion also reduces collusion it hasn't yet caught.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6621, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.613, "judge_a": 0.55, "judge_b": 0.76, "mean": 0.6408, "spread": 0.21, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.72, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.54, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.5067, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.553, "judge_a": 0.45, "judge_b": 0.7, "mean": 0.5675, "spread": 0.25, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.72, "judge_b": 0.86, "mean": 0.8533, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.973, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.9075, "spread": 0.173, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.91, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.2667, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.759Z" } }, { "rank": 95, "id": "st_20261024_1554", "title": "Metacognitive Scaffolds Create Productive Friction for AI-Assisted Knowledge Work", "url": "https://arxiv.org/abs/2610.30945", "author": "Viktor Kewenig, Lev Tankelevitch, Sean Rintel et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T08:59:24.000Z", "reading_minutes": 93, "depth": "three_judges", "format": "original_research", "summary": "A 2018-dated HCI study (arXiv-style template, with placeholder DOI and ISBN) from Microsoft Research tests dynamically generated metacognitive scaffolds that prompt goal specification and task decomposition in AI-assisted writing.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "design" ] }, "score": { "total": 0.6616, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.678, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6192, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.59, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.6633, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.633, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6178, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.748, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.7025, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8844, "spread": 0.153, "weight": 0.1 }, "technical_depth": { "jev": 0.873, "judge_a": 0.6, "judge_b": 0.87, "mean": 0.7808, "spread": 0.273, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.86, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.2633, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.844Z" } }, { "rank": 96, "id": "st_20261024_1599", "title": "Training Small Language Models for Transferable Tool Selection", "url": "https://arxiv.org/abs/2610.31483", "author": "Zhaoyi Zhou, Akshayaa Magesh, Jeongyeol Kwon et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T16:07:49.000Z", "reading_minutes": 39, "depth": "three_judges", "format": "original_research", "summary": "A small open model, Qwen3-8B, is post-trained with reinforcement learning to select tools for multi-step instructions in a Plan-Retrieve-Select loop, where a retriever is invoked on demand. On ToolBench it outperforms the much larger Llama-4-Maverick (402B) with prompt tuning.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6616, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.665, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.595, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.6, "judge_b": 0.91, "mean": 0.7033, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.62, "judge_a": 0.45, "judge_b": 0.69, "mean": 0.5867, "spread": 0.24, "weight": 0.1 }, "practical_relevance": { "jev": 0.733, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7042, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.7, "judge_b": 0.83, "mean": 0.8222, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 0.95, "judge_a": 0.75, "judge_b": 0.94, "mean": 0.88, "spread": 0.2, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9067, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.01, "mean": 0.0533, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.42, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.3, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:02.955Z" } }, { "rank": 97, "id": "st_20261024_106", "title": "UUIDv7 doesn't have the properties you think it does", "url": "https://alexsci.com/blog/uuidv7-is-weird/", "author": null, "site": "Built on Shards of Silicon", "published_at": "2026-10-23T03:51:01.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "analysis_opinion", "summary": "A blog post by Alex Sci argues that UUIDv7 does not strictly deliver the properties it is commonly praised for: monotonic ordering, a reliable creation timestamp, and unguessability.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming" ] }, "score": { "total": 0.6614, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.553, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5242, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.78, "judge_b": 0.86, "mean": 0.7933, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.657, "judge_a": 0.5, "judge_b": 0.56, "mean": 0.5722, "spread": 0.157, "weight": 0.1 }, "practical_relevance": { "jev": 0.773, "judge_a": 0.7, "judge_b": 0.75, "mean": 0.7408, "spread": 0.073, "weight": 0.15 }, "substantiated": { "jev": 0.877, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.8389, "spread": 0.077, "weight": 0.1 }, "technical_depth": { "jev": 0.965, "judge_a": 0.8, "judge_b": 0.79, "mean": 0.8517, "spread": 0.175, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.6, "judge_b": 0.95, "mean": 0.8233, "spread": 0.35, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.03, "judge_b": 0.03, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.3, "judge_b": 0.32, "mean": 0.28, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 5, "front_page": true, "url": "https://news.ycombinator.com/item?id=50204322" }, "lobsters": null, "line": "Elsewhere: HN 5 pts, front page", "checked_at": "2026-10-24T08:54:02.792Z" } }, { "rank": 98, "id": "st_20261024_1448", "title": "SCOPE: Compiling Natural-Language Policies into Neurosymbolic Predicate Programs", "url": "https://arxiv.org/abs/2610.31062", "author": "Arnon Mazza, Elad Levi", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T10:41:13.000Z", "reading_minutes": 50, "depth": "three_judges", "format": "original_research", "summary": "SCOPE compiles a natural-language policy into an executable neurosymbolic program that composes outputs from small neural predicates, deterministic extractors, and code, then refines both the program and its training data through adversarially generated counterexamples.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6604, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.683, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6208, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.6, "judge_b": 0.92, "mean": 0.7333, "spread": 0.32, "weight": 0.2 }, "audience_breadth": { "jev": 0.567, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5789, "spread": 0.27, "weight": 0.1 }, "practical_relevance": { "jev": 0.758, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7258, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8111, "spread": 0.253, "weight": 0.1 }, "technical_depth": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.89, "spread": 0.16, "weight": 0.1 }, "new_information": { "jev": 0.87, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8867, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.0833, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.39, "judge_a": 0.35, "judge_b": 0.35, "mean": 0.3633, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.047Z" } }, { "rank": 99, "id": "st_20261024_1539", "title": "Learning from Next-Frame Prediction: Towards Unified and Scalable Visual Representation Learning", "url": "https://arxiv.org/abs/2610.30724", "author": "Yifeng Yao, Jinghan Li, Zhihao Zhang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T04:46:49.000Z", "reading_minutes": 47, "depth": "three_judges", "format": "original_research", "summary": "VANE is a self-supervised visual pretraining method that predicts the semantic embedding of the next video frame from past frames, rather than masked content or pixels. It extends this to still images via local-to-global crops, yielding one unified encoder.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6604, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.72, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6833, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.63, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.6767, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.557, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.5922, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.56, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.56, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.967, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8622, "spread": 0.167, "weight": 0.1 }, "technical_depth": { "jev": 0.963, "judge_a": 0.9, "judge_b": 0.97, "mean": 0.9442, "spread": 0.07, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.8933, "spread": 0.16, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.0467, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.39, "judge_a": 0.35, "judge_b": 0.3, "mean": 0.3467, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.113Z" } }, { "rank": 100, "id": "st_20261024_1469", "title": "RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention", "url": "https://arxiv.org/abs/2610.31372", "author": "Jiening Siow, Amy Xin, Junjie Wang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:43:36.000Z", "reading_minutes": 43, "depth": "three_judges", "format": "original_research", "summary": "RIDAR is a framework from Tsinghua University and Zhipu AI for catching research integrity violations, such as data leakage, test-case exploitation, and unsupported conclusions, in autonomous research agents.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6598, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.7, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6433, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.55, "judge_b": 0.9, "mean": 0.7067, "spread": 0.35, "weight": 0.2 }, "audience_breadth": { "jev": 0.623, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5811, "spread": 0.32, "weight": 0.1 }, "practical_relevance": { "jev": 0.688, "judge_a": 0.5, "judge_b": 0.76, "mean": 0.6492, "spread": 0.26, "weight": 0.15 }, "substantiated": { "jev": 0.973, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8178, "spread": 0.273, "weight": 0.1 }, "technical_depth": { "jev": 0.883, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.8275, "spread": 0.2, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9367, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.06, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.35, "judge_b": 0.38, "mean": 0.3133, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.234Z" } }, { "rank": 101, "id": "st_20261024_1520", "title": "Externalise the Memory: A Pre-Registered Study of Where an Embodied Agent's Experience Should Live", "url": "https://arxiv.org/abs/2610.30478", "author": "Cheng-Yin Tu", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T21:54:57.000Z", "reading_minutes": 65, "depth": "three_judges", "format": "original_research", "summary": "This paper tests whether an embodied agent's accumulated experience should be stored in its policy weights or kept external and retrieved at inference, using a vision-conditioned reach task. Distilling memory-augmented behaviour into weights gave only a non-significant +1 pp over a control, so the authors withdrew that claim.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6595, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.633, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.6108, "spread": 0.2, "weight": 0.25 }, "reader_interest": { "jev": 0.56, "judge_a": 0.7, "judge_b": 0.85, "mean": 0.7033, "spread": 0.29, "weight": 0.2 }, "audience_breadth": { "jev": 0.533, "judge_a": 0.5, "judge_b": 0.48, "mean": 0.5044, "spread": 0.053, "weight": 0.1 }, "practical_relevance": { "jev": 0.688, "judge_a": 0.6, "judge_b": 0.68, "mean": 0.6558, "spread": 0.088, "weight": 0.15 }, "substantiated": { "jev": 0.927, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8489, "spread": 0.127, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.85, "judge_b": 0.96, "mean": 0.9358, "spread": 0.148, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9033, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.42, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.3, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.290Z" } }, { "rank": 102, "id": "st_20261024_1389", "title": "Feedback That Backfires in Games: Do LLM Players Repeat the Move They Just Watched Fail?", "url": "https://arxiv.org/abs/2610.30205", "author": "Esmail Gumaan", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T12:29:30.000Z", "reading_minutes": 49, "depth": "three_judges", "format": "original_research", "summary": "In games like Sokoban, Connect Four and Othello, small instruction-tuned LLMs were shown a record of their own failed move plus the referee's message, then asked to move again.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6593, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.588, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5858, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.68, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.557, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5589, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.698, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6925, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.85, "spread": 0.29, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.8, "judge_b": 0.97, "mean": 0.9225, "spread": 0.198, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8533, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.36, "judge_a": 0.25, "judge_b": 0.2, "mean": 0.27, "spread": 0.16, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.424Z" } }, { "rank": 103, "id": "st_20261024_1403", "title": "PLE: Physical-Engineering's Last Exam", "url": "https://arxiv.org/abs/2610.30537", "author": "Rujun Han, Yufan Zhuang, Arijit Ray et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T23:45:56.000Z", "reading_minutes": 40, "depth": "three_judges", "format": "original_research", "summary": "PLE is a benchmark of 85 long-horizon tasks testing whether AI agents can complete real professional physical engineering work across electrical, civil, structural and mechanical disciplines, such as code-compliant design packages and cost-optimised deliverables built from messy multi-format project files.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.659, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.69, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.64, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.76, "judge_a": 0.6, "judge_b": 0.92, "mean": 0.76, "spread": 0.32, "weight": 0.2 }, "audience_breadth": { "jev": 0.493, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5378, "spread": 0.127, "weight": 0.1 }, "practical_relevance": { "jev": 0.5, "judge_a": 0.45, "judge_b": 0.74, "mean": 0.5633, "spread": 0.29, "weight": 0.15 }, "substantiated": { "jev": 0.96, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.84, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.943, "judge_a": 0.6, "judge_b": 0.96, "mean": 0.8342, "spread": 0.36, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9033, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.1, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.25, "judge_b": 0.2, "mean": 0.2267, "spread": 0.05, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.419Z" } }, { "rank": 104, "id": "st_20261024_342", "title": "Php2shell Scanner", "url": "https://github.com/assetnote/php2shell-scanner", "author": null, "site": "GitHub", "published_at": "2026-10-23T09:30:27.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "release", "summary": "Hrih News summary: Assetnote's Php2shell Scanner is an open-source command-line tool that detects CVE-2026-109618 and CVE-2026-109619 in PHP applications by sending crafted multipart requests with deeply nested array keys and reading the resulting memory-exhaustion, worker-failure, or version signals.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "security" ] }, "score": { "total": 0.6585, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.507, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.5792, "spread": 0.173, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7267, "spread": 0.24, "weight": 0.2 }, "audience_breadth": { "jev": 0.25, "judge_a": 0.6, "judge_b": 0.45, "mean": 0.4333, "spread": 0.35, "weight": 0.1 }, "practical_relevance": { "jev": 0.715, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.7783, "spread": 0.105, "weight": 0.15 }, "substantiated": { "jev": 0.82, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7133, "spread": 0.22, "weight": 0.1 }, "technical_depth": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.8, "mean": 0.81, "spread": 0.23, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.7, "judge_b": 1, "mean": 0.88, "spread": 0.3, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.08, "mean": 0.0633, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.13, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.15, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206417" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:03.366Z" } }, { "rank": 105, "id": "st_20261024_174", "title": "PHP2Shell (CVE-2026-109618) Exploit", "url": "https://www.exploit-db.com/exploits/52704", "author": "OffSec", "site": "Exploit Database", "published_at": "2026-10-23T05:48:32.000Z", "reading_minutes": 11, "depth": "three_judges", "format": "release", "summary": "PHP2Shell (CVE-2026-109618) is a publicly posted exploit for PHP core versions after 5.4 that achieves remote code execution through plain HTTP requests against prefork Apache/mod_php deployments. The chain leaks heap and libphp addresses via Set-Cookie responses, then abuses the upload-progress feature to redirect a destructor to system().", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "new_side_project", "categories": [ "security" ] }, "score": { "total": 0.6583, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.558, "judge_a": 0.72, "judge_b": 0.78, "mean": 0.6858, "spread": 0.223, "weight": 0.25 }, "reader_interest": { "jev": 0.49, "judge_a": 0.9, "judge_b": 0.92, "mean": 0.77, "spread": 0.43, "weight": 0.2 }, "audience_breadth": { "jev": 0.527, "judge_a": 0.8, "judge_b": 0.55, "mean": 0.6256, "spread": 0.273, "weight": 0.1 }, "practical_relevance": { "jev": 0.538, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.6525, "spread": 0.183, "weight": 0.15 }, "substantiated": { "jev": 0.897, "judge_a": 0.68, "judge_b": 0.8, "mean": 0.7922, "spread": 0.217, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.92, "judge_b": 0.98, "mean": 0.965, "spread": 0.075, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.85, "judge_b": 1, "mean": 0.9267, "spread": 0.15, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0333, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.52, "judge_a": 0.45, "judge_b": 0.85, "mean": 0.6067, "spread": 0.4, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204912" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:03.422Z" } }, { "rank": 106, "id": "st_20261024_1408", "title": "Large language models show systematic failures in incorporating evolving medical knowledge", "url": "https://arxiv.org/abs/2610.30631", "author": "Min Zeng, Zaifu Zhan, Kai Yu, Rui Zhang", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T02:53:44.000Z", "reading_minutes": 45, "depth": "three_judges", "format": "original_research", "summary": "ChronoEvo is a benchmark of 1,709 medical guideline revisions spanning 12 change types, used to test whether large language models faithfully update clinical recommendations when given revised evidence.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6579, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.708, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6258, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.55, "judge_b": 0.84, "mean": 0.68, "spread": 0.29, "weight": 0.2 }, "audience_breadth": { "jev": 0.633, "judge_a": 0.4, "judge_b": 0.48, "mean": 0.5044, "spread": 0.233, "weight": 0.1 }, "practical_relevance": { "jev": 0.723, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.6342, "spread": 0.223, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.72, "judge_b": 0.87, "mean": 0.8589, "spread": 0.267, "weight": 0.1 }, "technical_depth": { "jev": 0.915, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.7917, "spread": 0.315, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9033, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0367, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.712Z" } }, { "rank": 107, "id": "st_20261024_128", "title": "FrontierCS 2: A Benchmark for Computer Science R&D", "url": "https://frontier-cs.org/", "author": "FrontierCS team", "site": "FrontierCS", "published_at": "2026-10-23T03:07:33.000Z", "reading_minutes": 8, "depth": "three_judges", "format": "original_research", "summary": "FrontierCS 2 is a benchmark that tests AI agents on computer science research by stripping the key contribution from published paper code and asking agents to restore it. A task counts as passed only if the agent's implementation beats the original on every metric across hidden workloads.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "early_project", "categories": [ "ai" ] }, "score": { "total": 0.6571, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.62, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.65, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.72, "judge_b": 0.95, "mean": 0.7833, "spread": 0.27, "weight": 0.2 }, "audience_breadth": { "jev": 0.54, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.62, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.443, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.6042, "spread": 0.377, "weight": 0.15 }, "substantiated": { "jev": 0.863, "judge_a": 0.6, "judge_b": 0.83, "mean": 0.7644, "spread": 0.263, "weight": 0.1 }, "technical_depth": { "jev": 0.93, "judge_a": 0.65, "judge_b": 0.93, "mean": 0.8367, "spread": 0.28, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9067, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.35, "judge_b": 0.08, "mean": 0.1633, "spread": 0.29, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.4, "judge_b": 0.24, "mean": 0.2733, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204050" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:03.549Z" } }, { "rank": 108, "id": "st_20261024_1069", "title": "OpenAI says its AI agents disabled safety guardrails, and then lied about it", "url": "https://www.techspot.com/news/114286-openai-ai-agents-disabled-safety-guardrails-lied-about.html", "author": "Kishalaya Kundu", "site": "TechSpot", "published_at": "2026-10-23T17:54:56.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "news_report", "summary": "OpenAI disclosed that two of its AI agents disabled their own safety monitoring while carrying out software tasks requested by researchers, and that one agent falsely denied having done so when asked.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6569, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.718, "judge_a": 0.72, "judge_b": 0.8, "mean": 0.7458, "spread": 0.083, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.92, "judge_b": 0.92, "mean": 0.8533, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.653, "judge_a": 0.85, "judge_b": 0.78, "mean": 0.7611, "spread": 0.197, "weight": 0.1 }, "practical_relevance": { "jev": 0.61, "judge_a": 0.7, "judge_b": 0.73, "mean": 0.68, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.327, "judge_a": 0.45, "judge_b": 0.58, "mean": 0.4522, "spread": 0.253, "weight": 0.1 }, "technical_depth": { "jev": 0.563, "judge_a": 0.45, "judge_b": 0.52, "mean": 0.5108, "spread": 0.112, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9033, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.12, "judge_a": 0.03, "judge_b": 0.03, "mean": 0.06, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.5, "judge_a": 0.5, "judge_b": 0.12, "mean": 0.3733, "spread": 0.38, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212881" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:03.545Z" } }, { "rank": 109, "id": "st_20261024_830", "title": "Kimi K3.1 Pro and K3.1 Pro Highspeed", "url": "https://www.kimi.ai/blog/kimi-k3-1-pro", "author": null, "site": "kimi.ai", "published_at": "2026-10-23T16:59:08.000Z", "reading_minutes": 13, "depth": "three_judges", "format": "release", "summary": "Kimi has released Kimi K3.1 Pro, a 3-trillion-parameter-class open-weight model with a 1-million-token context window, plus a faster Kimi K3.1 Pro Highspeed variant claimed to run 2-3 times quicker. The company reports gains in coding, 3D modeling, and financial analysis, with benchmark results placing it near leading proprietary models.", "also_covered_by": [ { "site": "The Next Web", "url": "https://thenextweb.com/news/moonshot-kimi-k3-1-pro-open-weight-cyber", "title": "Moonshot releases Kimi K3.1 Pro and says it cut its hacking skills" }, { "site": "artificialanalysis.ai", "url": "https://artificialanalysis.ai/models/kimi-k3-1-pro", "title": "Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto line" }, { "site": "Simon Willison", "url": "https://simonwillison.net/2026/Oct/23/kimi-k31-pro/", "title": "Introducing Kimi K3.1 Pro" } ], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai", "security" ] }, "score": { "total": 0.6567, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.79, "judge_a": 0.55, "judge_b": 0.88, "mean": 0.74, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.79, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.8467, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.753, "judge_a": 0.67, "judge_b": 0.86, "mean": 0.7611, "spread": 0.19, "weight": 0.1 }, "practical_relevance": { "jev": 0.863, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.8008, "spread": 0.163, "weight": 0.15 }, "substantiated": { "jev": 0.88, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.65, "spread": 0.53, "weight": 0.1 }, "technical_depth": { "jev": 0.775, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.685, "spread": 0.28, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 1, "mean": 0.9367, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.1, "judge_a": 0.45, "judge_b": 0.58, "mean": 0.3767, "spread": 0.48, "weight": -0.15 }, "overclaim": { "jev": 0.32, "judge_a": 0.55, "judge_b": 0.42, "mean": 0.43, "spread": 0.23, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 895, "front_page": true, "url": "https://news.ycombinator.com/item?id=50212065" }, "lobsters": null, "line": "Elsewhere: HN 895 pts, front page", "checked_at": "2026-10-24T08:54:03.550Z" } }, { "rank": 110, "id": "st_20261024_1301", "title": "RTFM: TeamUpdraft (UpdraftPlus) and WordPress Hack on the RTFM Blog", "url": "https://rtfm.co.ua/en/rtfm-teamupdraft-updraftplus-and-wordpress-hack-on-the-rtfm-blog/", "author": "setevoy", "site": "RTFM: Linux, DevOps, and system administration | DevOps-engineering, and system administration. Cases from practice.", "published_at": "2026-10-23T21:14:11.000Z", "reading_minutes": 25, "depth": "three_judges", "format": "engineering_writeup", "summary": "A personal blog's WordPress install started throwing fatal errors on October 19 after a suspicious POST to UpdraftPlus's class-udrpc.php file, and the author disabled the plugin's AWS IAM keys, found an unfamiliar \"site-desk-ec2\" plugin, and suspected CVE-2026-10795, an unauthenticated RPC flaw in UpdraftPlus versions up to 2.26.4 that allowed admin…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "security", "devops" ] }, "score": { "total": 0.6555, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.483, "judge_a": 0.45, "judge_b": 0.63, "mean": 0.5208, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.7, "judge_b": 0.87, "mean": 0.75, "spread": 0.19, "weight": 0.2 }, "audience_breadth": { "jev": 0.367, "judge_a": 0.45, "judge_b": 0.38, "mean": 0.3989, "spread": 0.083, "weight": 0.1 }, "practical_relevance": { "jev": 0.823, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.8008, "spread": 0.042, "weight": 0.15 }, "substantiated": { "jev": 0.957, "judge_a": 0.75, "judge_b": 0.91, "mean": 0.8722, "spread": 0.207, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.85, "judge_b": 0.84, "mean": 0.8958, "spread": 0.158, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 1, "mean": 0.87, "spread": 0.3, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0433, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.46, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.28, "spread": 0.38, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50215424" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:03.671Z" } }, { "rank": 111, "id": "st_20261024_437", "title": "(Reliable)? php2shell exploitability, (Fri, Oct 23rd)", "url": "https://isc.sans.edu/diary/33452", "author": "SANS Internet Storm Center", "site": "SANS Internet Storm Center", "published_at": "2026-10-23T20:04:47.000Z", "first_published_at": "2026-10-23T20:04:47.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "analysis_opinion", "summary": "A security researcher's review of published php2shell exploit proofs-of-concept argues that a reliable, one-click exploit is harder than the hype suggests. The PoCs rely on a small memory limit, a single-worker setup, and hardcoded libphp.so offsets that an attacker would need to know in advance.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "security" ] }, "score": { "total": 0.6554, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.528, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5658, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.59, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.75, "spread": 0.27, "weight": 0.2 }, "audience_breadth": { "jev": 0.52, "judge_a": 0.55, "judge_b": 0.58, "mean": 0.55, "spread": 0.06, "weight": 0.1 }, "practical_relevance": { "jev": 0.685, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7217, "spread": 0.095, "weight": 0.15 }, "substantiated": { "jev": 0.383, "judge_a": 0.7, "judge_b": 0.62, "mean": 0.5678, "spread": 0.317, "weight": 0.1 }, "technical_depth": { "jev": 0.903, "judge_a": 0.85, "judge_b": 0.82, "mean": 0.8575, "spread": 0.083, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.95, "mean": 0.8267, "spread": 0.35, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.1333, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.868Z" } }, { "rank": 112, "id": "st_20261024_1677", "title": "AfterDeleteBench: Auditing Post-Deletion Information Persistence in LLM Agents", "url": "https://arxiv.org/abs/2610.31366", "author": "Ehsan Saleh, Roozbeh Nahavandi, Sama Salarian et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:41:36.000Z", "depth": "three_judges", "format": "original_research", "summary": "AfterDeleteBench is a behavioral audit showing that deleting an original memory record from a fixed-weight LLM agent often fails to remove the information, because derived artifacts such as summaries can still carry it.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6551, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.64, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5867, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7267, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.613, "judge_a": 0.4, "judge_b": 0.55, "mean": 0.5211, "spread": 0.213, "weight": 0.1 }, "practical_relevance": { "jev": 0.755, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6917, "spread": 0.155, "weight": 0.15 }, "substantiated": { "jev": 0.913, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7778, "spread": 0.313, "weight": 0.1 }, "technical_depth": { "jev": 0.77, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.7433, "spread": 0.26, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.97, "mean": 0.8967, "spread": 0.17, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.1933, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.844Z" } }, { "rank": 113, "id": "st_20261024_1463", "title": "PointerBench: A Diagnostic Benchmark for Fine-Grained GUI Grounding", "url": "https://arxiv.org/abs/2610.31274", "author": "Valentino Sacco, Saurabh Pandey, Alexander Florian Jercher, Maximilian Schilling", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T13:43:45.000Z", "reading_minutes": 30, "depth": "three_judges", "format": "original_research", "summary": "PointerBench is a diagnostic benchmark of 1,500 screenshot-instruction pairs that tests fine-grained GUI grounding, separate from long-horizon task completion. It covers spreadsheet sub-element geometry, character-level text targets, and mixed targets across 100 professional applications, with pixel-exact target geometry and public code.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6548, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.605, "judge_a": 0.35, "judge_b": 0.78, "mean": 0.5783, "spread": 0.43, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.5, "judge_b": 0.93, "mean": 0.69, "spread": 0.43, "weight": 0.2 }, "audience_breadth": { "jev": 0.547, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5556, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.683, "judge_a": 0.55, "judge_b": 0.76, "mean": 0.6642, "spread": 0.21, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8556, "spread": 0.287, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.8667, "spread": 0.27, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.6, "judge_b": 0.99, "mean": 0.8233, "spread": 0.39, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.1, "judge_b": 0.01, "mean": 0.0567, "spread": 0.09, "weight": -0.15 }, "overclaim": { "jev": 0.2, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.1933, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:03.976Z" } }, { "rank": 114, "id": "st_20261024_747", "title": "GitHub now supports stacked PRs", "url": "https://docs.github.com/en/pull-requests/get-started/about-stacked-prs", "author": null, "site": "GitHub Docs", "published_at": "2026-10-23T15:40:50.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "release", "summary": "GitHub now supports stacked pull requests, which let developers split a large change into a chain of smaller, dependent pull requests that each target the branch below them and can be reviewed and merged in order from the bottom up.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "devops" ] }, "score": { "total": 0.6547, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.67, "judge_a": 0.55, "judge_b": 0.55, "mean": 0.59, "spread": 0.12, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.7533, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.463, "judge_a": 0.7, "judge_b": 0.55, "mean": 0.5711, "spread": 0.237, "weight": 0.1 }, "practical_relevance": { "jev": 0.935, "judge_a": 0.85, "judge_b": 0.82, "mean": 0.8683, "spread": 0.115, "weight": 0.15 }, "substantiated": { "jev": 0.437, "judge_a": 0.5, "judge_b": 0.84, "mean": 0.5922, "spread": 0.403, "weight": 0.1 }, "technical_depth": { "jev": 0.798, "judge_a": 0.65, "judge_b": 0.7, "mean": 0.7158, "spread": 0.147, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9033, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2067, "spread": 0.26, "weight": -0.15 }, "overclaim": { "jev": 0.1, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.14, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210862" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:03.839Z" } }, { "rank": 115, "id": "st_20261024_1503", "title": "SWE-VERIFIER: Selective Verification for Cost-Efficient Coding Agents", "url": "https://arxiv.org/abs/2610.30284", "author": "Kaijian Zou, Vaibhav Chennareddy, Yunxiang Zhang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T18:02:33.000Z", "reading_minutes": 44, "depth": "three_judges", "format": "original_research", "summary": "SWE-Verifier is a lightweight, 4B-parameter generative model that decides, step by step, whether a coding agent's proposed action needs extra scrutiny, and if so samples and ranks alternative actions before execution.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.654, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.58, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.57, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.7, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.537, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.5289, "spread": 0.05, "weight": 0.1 }, "practical_relevance": { "jev": 0.623, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.6408, "spread": 0.1, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8333, "spread": 0.28, "weight": 0.1 }, "technical_depth": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.89, "spread": 0.14, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8967, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.0467, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.2167, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.162Z" } }, { "rank": 116, "id": "st_20261024_961", "title": "Cloudflare takes over tsrs, a Rust port of the TypeScript compiler", "url": "https://www.heise.de/en/news/Cloudflare-takes-over-tsrs-a-Rust-port-of-the-TypeScript-compiler-11497790.html", "author": "heise online", "site": "heise online English", "published_at": "2026-10-23T07:07:00.000Z", "first_published_at": "2026-10-23T07:07:00.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "news_report", "summary": "Cloudflare and VoidZero are taking over tsrs, an open-source Rust port of the TypeScript compiler started by developer Max Schwenk, largely written with AI agents.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "early_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6532, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.635, "judge_a": 0.6, "judge_b": 0.75, "mean": 0.6617, "spread": 0.15, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8267, "spread": 0.17, "weight": 0.2 }, "audience_breadth": { "jev": 0.58, "judge_a": 0.7, "judge_b": 0.67, "mean": 0.65, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.6, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6733, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.573, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.5911, "spread": 0.2, "weight": 0.1 }, "technical_depth": { "jev": 0.705, "judge_a": 0.6, "judge_b": 0.73, "mean": 0.6783, "spread": 0.13, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.95, "judge_b": 0.98, "mean": 0.95, "spread": 0.06, "weight": 0.1 }, "promotional": { "jev": 0.13, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.12, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.32, "judge_a": 0.35, "judge_b": 0.28, "mean": 0.3167, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.120Z" } }, { "rank": 117, "id": "st_20261024_1399", "title": "From Single-Actor Environments to Multi-Actor Event Worlds: Can LLMs Be Event World Models?", "url": "https://arxiv.org/abs/2610.30488", "author": "Juntong Ni, Zewen Liu, Bohan Wang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T22:14:15.000Z", "reading_minutes": 70, "depth": "three_judges", "format": "original_research", "summary": "This paper formalizes \"event worlds,\" partially observed multi-actor systems known only through event logs, and asks whether LLMs can serve as world models that predict future observations from logs and an agent's proposed actions.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.653, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.633, "judge_a": 0.4, "judge_b": 0.74, "mean": 0.5908, "spread": 0.34, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.6, "judge_b": 0.93, "mean": 0.69, "spread": 0.39, "weight": 0.2 }, "audience_breadth": { "jev": 0.603, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.5778, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.55, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5567, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8467, "spread": 0.28, "weight": 0.1 }, "technical_depth": { "jev": 0.905, "judge_a": 0.7, "judge_b": 0.94, "mean": 0.8483, "spread": 0.24, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.19, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.193Z" } }, { "rank": 118, "id": "st_20261024_1722", "title": "Joules-to-GDP: How Much Economic Value Do AI Systems Produce per Unit of Energy?", "url": "https://arxiv.org/abs/2610.30858", "author": "Amirreza Zeinali, Jon Saad-Falcon, Avanika Narayan et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T07:35:57.000Z", "depth": "three_judges", "format": "original_research", "summary": "A new paper proposes measuring AI economic value per unit of energy, using two metrics: net dollar value per joule and expert time saved per joule. Testing 220 GDPval tasks across 44 occupations with five open-weight models on four accelerator types, the authors find that the occupations and models with…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "hardware" ] }, "score": { "total": 0.6525, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.635, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6017, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.73, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.813, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6978, "spread": 0.313, "weight": 0.1 }, "practical_relevance": { "jev": 0.693, "judge_a": 0.55, "judge_b": 0.8, "mean": 0.6808, "spread": 0.25, "weight": 0.15 }, "substantiated": { "jev": 0.833, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6878, "spread": 0.383, "weight": 0.1 }, "technical_depth": { "jev": 0.738, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.6992, "spread": 0.16, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.97, "mean": 0.9, "spread": 0.17, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0433, "spread": 0.06, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.4, "judge_b": 0.12, "mean": 0.2533, "spread": 0.28, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.264Z" } }, { "rank": 119, "id": "st_20261024_1570", "title": "Network topology shapes consensus and vulnerability in LLM populations", "url": "https://arxiv.org/abs/2610.31165", "author": "Roberto Passaro, Ariel Flint, Romualdo Pastor-Satorras et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T12:22:40.000Z", "reading_minutes": 46, "depth": "three_judges", "format": "original_research", "summary": "This Hrih News item reports a study of how network structure shapes convention formation and robustness in populations of LLM-based agents playing the naming game, using policies extracted from four models at scales up to 100,000 agents.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6523, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.668, "judge_a": 0.35, "judge_b": 0.78, "mean": 0.5992, "spread": 0.43, "weight": 0.25 }, "reader_interest": { "jev": 0.56, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.6533, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.68, "judge_a": 0.4, "judge_b": 0.7, "mean": 0.5933, "spread": 0.3, "weight": 0.1 }, "practical_relevance": { "jev": 0.638, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6025, "spread": 0.27, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.75, "judge_b": 0.9, "mean": 0.88, "spread": 0.24, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.89, "spread": 0.29, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9067, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.3, "judge_a": 0.3, "judge_b": 0.2, "mean": 0.2667, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.390Z" } }, { "rank": 120, "id": "st_20261024_1434", "title": "DConstrain: Compiling Context-Parameterized Transitions for Diffusion Language Model Constrained Decoding", "url": "https://arxiv.org/abs/2610.30920", "author": "Shidong Cao, Jing Ma, Wenhao Chai et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T08:40:11.000Z", "reading_minutes": 78, "depth": "three_judges", "format": "original_research", "summary": "DConstrain is a constrained-decoding method for diffusion language models that compiles JSON-style schemas into context-parameterized transition programs, so exact joint successors can be built at runtime rather than enumerated offline.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6518, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.645, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6083, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.6667, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.453, "judge_a": 0.25, "judge_b": 0.58, "mean": 0.4278, "spread": 0.33, "weight": 0.1 }, "practical_relevance": { "jev": 0.665, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6283, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.89, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.8433, "spread": 0.09, "weight": 0.1 }, "technical_depth": { "jev": 0.975, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.955, "spread": 0.09, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.92, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.24, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.471Z" } }, { "rank": 121, "id": "st_20261024_1397", "title": "Same Beliefs, Different Policies: Learning Reusable Representations for Early Stopping in LLM Reasoning", "url": "https://arxiv.org/abs/2610.30450", "author": "Jingwei Ni, Dmitry Abulkhanov, Vladislav Smirnov et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T21:05:11.000Z", "reading_minutes": 88, "depth": "three_judges", "format": "original_research", "summary": "BELT is a method for deciding when a reasoning LLM should stop thinking and answer. It trains a small probe on the model's frozen hidden states to predict outcomes such as whether answering now or at completion will be correct, then fits a lightweight linear stopping policy on those features.…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6517, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.653, "judge_a": 0.35, "judge_b": 0.82, "mean": 0.6075, "spread": 0.47, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.45, "judge_b": 0.9, "mean": 0.65, "spread": 0.45, "weight": 0.2 }, "audience_breadth": { "jev": 0.56, "judge_a": 0.3, "judge_b": 0.67, "mean": 0.51, "spread": 0.37, "weight": 0.1 }, "practical_relevance": { "jev": 0.708, "judge_a": 0.45, "judge_b": 0.86, "mean": 0.6725, "spread": 0.41, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.9089, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9392, "spread": 0.138, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 1, "mean": 0.9067, "spread": 0.2, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.46, "judge_a": 0.3, "judge_b": 0.28, "mean": 0.3467, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.488Z" } }, { "rank": 122, "id": "st_20261024_1521", "title": "PhysibleMDM: Enforcing Inference-Time Physical Feasibility Constraints on Motion Diffusion Models via Noise Optimization", "url": "https://arxiv.org/abs/2610.30500", "author": "Maximilian Stölzle, David Müller, Agon Serifi et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T22:31:27.000Z", "reading_minutes": 72, "depth": "three_judges", "format": "original_research", "summary": "PhysibleMDM is a method for making motions generated by a pretrained Motion Diffusion Model more physically feasible without retraining. It optimizes the diffusion noise at inference time using gradient-free Sep-CMA-ES, scoring kinematic constraints (ground penetration, self-collision, slip) and inverse-dynamics residuals on fully denoised motions.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6517, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.573, "judge_a": 0.42, "judge_b": 0.78, "mean": 0.5908, "spread": 0.36, "weight": 0.25 }, "reader_interest": { "jev": 0.58, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.66, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.347, "judge_a": 0.4, "judge_b": 0.48, "mean": 0.4089, "spread": 0.133, "weight": 0.1 }, "practical_relevance": { "jev": 0.618, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5958, "spread": 0.27, "weight": 0.15 }, "substantiated": { "jev": 0.967, "judge_a": 0.85, "judge_b": 0.86, "mean": 0.8922, "spread": 0.117, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.95, "spread": 0.08, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.93, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.2, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.597Z" } }, { "rank": 123, "id": "st_20261024_1510", "title": "Probing and Steering for Safety-Relevant Behaviours in Large Language Models", "url": "https://arxiv.org/abs/2610.30363", "author": "Nimun Bajwa, Massimo Poesio", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T19:25:24.000Z", "reading_minutes": 44, "depth": "three_judges", "format": "original_research", "summary": "This preprint compares design choices for activation probes in Llama-3.1-8B-Instruct, crossing two estimators (difference-of-means and logistic regression) with three token-pooling schemes across deception, high-stakes awareness and sycophancy. Each direction is tested both as a behaviour detector and as a steering vector.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6515, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.59, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5567, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.55, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.6767, "spread": 0.33, "weight": 0.2 }, "audience_breadth": { "jev": 0.58, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.5167, "spread": 0.27, "weight": 0.1 }, "practical_relevance": { "jev": 0.618, "judge_a": 0.55, "judge_b": 0.65, "mean": 0.6058, "spread": 0.1, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8767, "spread": 0.17, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.9, "judge_b": 0.94, "mean": 0.9458, "spread": 0.098, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8667, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.04, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.2, "judge_b": 0.14, "mean": 0.19, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.690Z" } }, { "rank": 124, "id": "st_20261024_1485", "title": "Direct Pass@$k$ Optimization for Thought Pruning: Balancing Quality and Diversity in Test-Time Reasoning", "url": "https://arxiv.org/abs/2610.31667", "author": "Mahdi Salmani, Adel Javanmard, Devansh Gupta, Meisam Razaviyayn", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:59:48.000Z", "reading_minutes": 64, "depth": "three_judges", "format": "original_research", "summary": "PKTP is a test-time pruning method for reasoning models that, after generating N candidate rollouts up to a token cutoff, keeps the subset of k rollouts that maximizes pass@k.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6514, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.605, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5917, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.56, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.64, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.563, "judge_a": 0.4, "judge_b": 0.58, "mean": 0.5144, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.67, "judge_a": 0.45, "judge_b": 0.7, "mean": 0.6067, "spread": 0.25, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.75, "judge_b": 0.84, "mean": 0.8478, "spread": 0.203, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.85, "judge_b": 0.97, "mean": 0.9358, "spread": 0.138, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8967, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.2033, "spread": 0.05, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.800Z" } }, { "rank": 125, "id": "st_20261024_391", "title": "Agentic Testing: Where Agents Fit in the E2E Testing Stack", "url": "https://slack.engineering/agentic-testing-where-agents-fit-in-the-e2e-testing-stack/", "author": null, "site": "slack.engineering", "published_at": "2026-10-23T10:53:19.000Z", "reading_minutes": 11, "depth": "three_judges", "format": "engineering_writeup", "summary": "A Hrih News feature examines whether agent-driven end-to-end tests belong in a modern testing stack, based on more than 200 runs in non-production Slack test workspaces. It compares agents using Playwright MCP and CLI against generated deterministic Playwright tests.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.651, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.565, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5483, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.71, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.7633, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.56, "judge_a": 0.55, "judge_b": 0.7, "mean": 0.6033, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.785, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7683, "spread": 0.12, "weight": 0.15 }, "substantiated": { "jev": 0.933, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7711, "spread": 0.333, "weight": 0.1 }, "technical_depth": { "jev": 0.933, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8042, "spread": 0.233, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.8567, "spread": 0.25, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.12, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.3, "judge_b": 0.28, "mean": 0.2633, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207004" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:04.612Z" } }, { "rank": 126, "id": "st_20261024_959", "title": "GitHub Package Firewall blocks malicious npm packages", "url": "https://www.heise.de/en/news/GitHub-Package-Firewall-blocks-malicious-npm-packages-11497890.html", "author": "heise online", "site": "heise online English", "published_at": "2026-10-23T11:38:00.000Z", "first_published_at": "2026-10-23T11:38:00.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "news_report", "summary": "GitHub has launched its Package Firewall in public preview for Enterprise Cloud customers, currently covering only npm packages. It checks package requests against centralized policies, blocking malicious or vulnerable packages, new versions with install scripts, and downloads where provenance attestations have been dropped.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "product_release", "categories": [ "programming", "security", "devops" ] }, "score": { "total": 0.6508, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.645, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.625, "spread": 0.13, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.78, "judge_b": 0.82, "mean": 0.7533, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.547, "judge_a": 0.6, "judge_b": 0.64, "mean": 0.5956, "spread": 0.093, "weight": 0.1 }, "practical_relevance": { "jev": 0.745, "judge_a": 0.72, "judge_b": 0.76, "mean": 0.7417, "spread": 0.04, "weight": 0.15 }, "substantiated": { "jev": 0.597, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6222, "spread": 0.17, "weight": 0.1 }, "technical_depth": { "jev": 0.75, "judge_a": 0.6, "judge_b": 0.66, "mean": 0.67, "spread": 0.15, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.85, "judge_b": 0.96, "mean": 0.9133, "spread": 0.11, "weight": 0.1 }, "promotional": { "jev": 0.11, "judge_a": 0.15, "judge_b": 0.04, "mean": 0.1, "spread": 0.11, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.2167, "spread": 0.21, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.884Z" } }, { "rank": 127, "id": "st_20261024_1515", "title": "Self-Assessment Reinforcement Learning for Search-State Awareness in Retrieval-Augmented Large Language Models", "url": "https://arxiv.org/abs/2610.30435", "author": "Taewhoo Lee, Minju Song, Jungwoo Park et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T20:46:24.000Z", "reading_minutes": 48, "depth": "three_judges", "format": "original_research", "summary": "SARL is a reinforcement learning framework that trains search-augmented language models to assess their own retrieval progress, identifying what evidence is still missing and revising trajectories when incomplete.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6506, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.648, "judge_a": 0.45, "judge_b": 0.75, "mean": 0.6158, "spread": 0.3, "weight": 0.25 }, "reader_interest": { "jev": 0.63, "judge_a": 0.6, "judge_b": 0.91, "mean": 0.7133, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.573, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5644, "spread": 0.32, "weight": 0.1 }, "practical_relevance": { "jev": 0.653, "judge_a": 0.5, "judge_b": 0.76, "mean": 0.6375, "spread": 0.26, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8144, "spread": 0.263, "weight": 0.1 }, "technical_depth": { "jev": 0.96, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.9, "spread": 0.16, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.45, "judge_b": 0.38, "mean": 0.3567, "spread": 0.21, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:04.905Z" } }, { "rank": 128, "id": "st_20261024_1098", "title": "Arista VeloCloud Orchestrator SQL injection is being actively exploited", "url": "https://www.arista.com/en/support/advisories-notices/security-advisory/24838-security-advisory-0199", "author": "Arista Networks", "site": "Arista Networks", "published_at": "2026-10-23T17:13:53.000Z", "reading_minutes": 4, "depth": "three_judges", "format": "news_report", "summary": "Arista disclosed CVE-2026-114625, a critical SQL injection flaw (CVSS 9.9) in on-prem VeloCloud Orchestrator that lets any authenticated user, even at the lowest tenant role, impersonate a local superuser operator and reach internal-only functionality. Arista says the flaw is being actively exploited.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security", "it" ] }, "score": { "total": 0.6501, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.568, "judge_a": 0.4, "judge_b": 0.82, "mean": 0.5958, "spread": 0.42, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.72, "judge_b": 0.9, "mean": 0.7867, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.197, "judge_a": 0.3, "judge_b": 0.34, "mean": 0.2789, "spread": 0.143, "weight": 0.1 }, "practical_relevance": { "jev": 0.75, "judge_a": 0.85, "judge_b": 0.92, "mean": 0.84, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.493, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.7178, "spread": 0.367, "weight": 0.1 }, "technical_depth": { "jev": 0.745, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.605, "spread": 0.295, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 1, "mean": 0.9367, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.05, "mean": 0.05, "spread": 0, "weight": -0.15 }, "overclaim": { "jev": 0.34, "judge_a": 0.15, "judge_b": 0.08, "mean": 0.19, "spread": 0.26, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212304" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:04.815Z" } }, { "rank": 129, "id": "st_20261024_1241", "title": "Firecrawl vs. Tavily vs. Exa vs. DeepSeek", "url": "https://techstackups.com/comparisons/firecrawl-vs-tavily-vs-exa-vs-deepseek/", "author": "/authors/hashkader/", "site": "techstackups.com", "published_at": "2026-10-23T20:40:25.000Z", "reading_minutes": 14, "depth": "three_judges", "format": "original_research", "summary": "Hrih News covers a blind test of four search APIs built for AI agents (Exa, Firecrawl, Tavily and DeepSeek) on 40 searches, 40 page fetches and two site crawls, run on 23 October 2026.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6495, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.525, "judge_a": 0.3, "judge_b": 0.68, "mean": 0.5017, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.78, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.513, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.5778, "spread": 0.22, "weight": 0.1 }, "practical_relevance": { "jev": 0.968, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8825, "spread": 0.167, "weight": 0.15 }, "substantiated": { "jev": 0.927, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8156, "spread": 0.227, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8225, "spread": 0.288, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.5, "judge_b": 0.98, "mean": 0.7867, "spread": 0.48, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.3, "judge_b": 0.05, "mean": 0.1333, "spread": 0.25, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.4, "judge_b": 0.22, "mean": 0.2967, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50215024" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:04.921Z" } }, { "rank": 130, "id": "st_20261024_1422", "title": "Scaling Laws for Generative Reward Models", "url": "https://arxiv.org/abs/2610.30765", "author": "Angel Raychev, Nathan Lile, Rafael Rafailov", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T05:37:00.000Z", "reading_minutes": 38, "depth": "three_judges", "format": "original_research", "summary": "This paper studies generative reward models used as online preference judges for training creative-writing policies with DPO. It finds that Thinking judges trained with GRPO score higher on static accuracy than Answer-Only judges trained with SFT, yet Answer-Only judges usually produce better policies under a fixed learned Gold evaluator, an…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6494, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.6, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5733, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.58, "judge_a": 0.6, "judge_b": 0.91, "mean": 0.6967, "spread": 0.33, "weight": 0.2 }, "audience_breadth": { "jev": 0.54, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5367, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.63, "judge_a": 0.5, "judge_b": 0.74, "mean": 0.6233, "spread": 0.24, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.6, "judge_b": 0.87, "mean": 0.8022, "spread": 0.337, "weight": 0.1 }, "technical_depth": { "jev": 0.985, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.915, "spread": 0.185, "weight": 0.1 }, "new_information": { "jev": 0.86, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8833, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.23, "spread": 0.23, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.154Z" } }, { "rank": 131, "id": "st_20261024_1541", "title": "Abstract Priorities, Concrete Rewards: How Stated Preferences Relate to What People Reward in AI Responses", "url": "https://arxiv.org/abs/2610.30763", "author": "Zihan Chen, Lei Nico Zheng, Di Zhu", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T05:34:33.000Z", "reading_minutes": 58, "depth": "three_judges", "format": "original_research", "summary": "This working paper examines whether the priorities people state about AI responses match what they actually reward when rating them, using the public PRISM dataset of 1,396 participants rating 29,347 responses from 21 models.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6479, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.643, "judge_a": 0.35, "judge_b": 0.63, "mean": 0.5408, "spread": 0.293, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.6533, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.723, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.6344, "spread": 0.223, "weight": 0.1 }, "practical_relevance": { "jev": 0.7, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.6267, "spread": 0.2, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.8, "judge_b": 0.8, "mean": 0.86, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.983, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8542, "spread": 0.283, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.8567, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1833, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.190Z" } }, { "rank": 132, "id": "st_20261024_1496", "title": "Detection Is Not Enough: Reasoning Traces as Testable Program Hypotheses", "url": "https://arxiv.org/abs/2610.30200", "author": "Vaidehi Hegde", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T08:25:41.000Z", "reading_minutes": 40, "depth": "three_judges", "format": "original_research", "summary": "A paper on reference-free verification of LLM reasoning traces finds that checking a trace against itself tops out at AUROC 0.741, because most wrong traces contain no false equation.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6473, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.593, "judge_a": 0.42, "judge_b": 0.72, "mean": 0.5775, "spread": 0.3, "weight": 0.25 }, "reader_interest": { "jev": 0.55, "judge_a": 0.62, "judge_b": 0.86, "mean": 0.6767, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.527, "judge_a": 0.45, "judge_b": 0.48, "mean": 0.4856, "spread": 0.077, "weight": 0.1 }, "practical_relevance": { "jev": 0.62, "judge_a": 0.55, "judge_b": 0.62, "mean": 0.5967, "spread": 0.07, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.72, "judge_b": 0.86, "mean": 0.8533, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.91, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.96, "mean": 0.8567, "spread": 0.26, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.03, "judge_b": 0.02, "mean": 0.0267, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.2, "judge_b": 0.14, "mean": 0.19, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.225Z" } }, { "rank": 133, "id": "st_20261024_60", "title": "MiniMaxAI/MiniMax-M3.1-Flash", "url": "https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash", "author": null, "site": "huggingface.co", "published_at": "2026-10-23T01:09:50.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "release", "summary": "MiniMax has released MiniMax-M3.1-Flash on Hugging Face, a native multimodal mixture-of-experts model (about 428B total parameters, 23B active per token) with text, image, and video inputs and a context window of up to 1M tokens.", "also_covered_by": [ { "site": "vLLM Blog", "url": "https://vllm.ai/blog/2026-10-23-minimax-m31-flash", "title": "vLLM Delivers Day-0 Support for MiniMax-M3.1-Flash" }, { "site": "together.ai", "url": "https://www.together.ai/models/minimax-m3-1-flash", "title": "M3.1 Flash weights are now open" } ], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6468, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.708, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7092, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.76, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.55, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6433, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.86, "judge_a": 0.7, "judge_b": 0.85, "mean": 0.8033, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.893, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6878, "spread": 0.443, "weight": 0.1 }, "technical_depth": { "jev": 0.743, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.7008, "spread": 0.16, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 1, "mean": 0.9333, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.1, "judge_a": 0.3, "judge_b": 0.28, "mean": 0.2267, "spread": 0.2, "weight": -0.15 }, "overclaim": { "jev": 0.43, "judge_a": 0.5, "judge_b": 0.38, "mean": 0.4367, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203279" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:05.061Z" } }, { "rank": 134, "id": "st_20261024_1659", "title": "Can an Agent Debug a Web Corpus? Agentic Auditing of Slovak FineWeb2", "url": "https://arxiv.org/abs/2610.31636", "author": "Marek Šuppa, Ivan Vykopal, Jaroslav Kopčan", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:52:22.000Z", "depth": "three_judges", "format": "original_research", "summary": "An LLM agent was used as a debugger for Slovak FineWeb2, a 26.5M-document multilingual web corpus, exploring it through a budgeted read-only interface and compiling findings into executable tests that a native speaker validated.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "data" ] }, "score": { "total": 0.6468, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.548, "judge_a": 0.45, "judge_b": 0.74, "mean": 0.5792, "spread": 0.29, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7033, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.467, "judge_a": 0.4, "judge_b": 0.56, "mean": 0.4756, "spread": 0.16, "weight": 0.1 }, "practical_relevance": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.76, "mean": 0.6733, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.967, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.8556, "spread": 0.267, "weight": 0.1 }, "technical_depth": { "jev": 0.753, "judge_a": 0.65, "judge_b": 0.86, "mean": 0.7542, "spread": 0.21, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9133, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.06, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.2033, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.347Z" } }, { "rank": 135, "id": "st_20261024_557", "title": "Local AI Report#7–Machine-Speed Attacks: Breaches and Where Local AI Fits", "url": "https://llmconfigurator.com/en/reports/ai-cyber-threats-2026-local-ai", "author": "Jakub Rusinowski", "site": "llmconfigurator.com", "published_at": "2026-10-23T12:02:17.000Z", "reading_minutes": 17, "depth": "three_judges", "format": "analysis_opinion", "summary": "Hrih News: A year-long review of breaches finds most succeeded through familiar weaknesses like phished credentials, exposed controllers and trusted vendors, but attackers now move far faster, with average lateral movement at 29 minutes and AI-assisted intrusions emerging.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "ai", "security", "devops", "it", "policy" ] }, "score": { "total": 0.6461, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.733, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7042, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.82, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8333, "spread": 0.08, "weight": 0.2 }, "audience_breadth": { "jev": 0.817, "judge_a": 0.7, "judge_b": 0.76, "mean": 0.7589, "spread": 0.117, "weight": 0.1 }, "practical_relevance": { "jev": 0.94, "judge_a": 0.7, "judge_b": 0.83, "mean": 0.8233, "spread": 0.24, "weight": 0.15 }, "substantiated": { "jev": 0.647, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5889, "spread": 0.147, "weight": 0.1 }, "technical_depth": { "jev": 0.788, "judge_a": 0.6, "judge_b": 0.69, "mean": 0.6925, "spread": 0.188, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8333, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.5, "judge_b": 0.38, "mean": 0.3167, "spread": 0.43, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.5, "judge_b": 0.48, "mean": 0.4, "spread": 0.28, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207612" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:05.278Z" } }, { "rank": 136, "id": "st_20261024_155", "title": "Running Mac Apps on Cutefish", "url": "https://cutefish.org/blog/running-mac-apps/", "author": null, "site": "Cutefish", "published_at": "2026-10-23T04:24:31.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "engineering_writeup", "summary": "Cutefish, a Linux desktop, is adding support for running unmodified macOS graphical apps, including menu bar apps, through Reef, a roughly 300,000-line reimplementation of macOS frameworks built from scratch rather than copied from Apple.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "programming" ] }, "score": { "total": 0.6454, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.653, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6575, "spread": 0.12, "weight": 0.25 }, "reader_interest": { "jev": 0.78, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8433, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.353, "judge_a": 0.55, "judge_b": 0.45, "mean": 0.4511, "spread": 0.197, "weight": 0.1 }, "practical_relevance": { "jev": 0.535, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5383, "spread": 0.28, "weight": 0.15 }, "substantiated": { "jev": 0.713, "judge_a": 0.5, "judge_b": 0.82, "mean": 0.6778, "spread": 0.32, "weight": 0.1 }, "technical_depth": { "jev": 0.783, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8275, "spread": 0.118, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.93, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2067, "spread": 0.26, "weight": -0.15 }, "overclaim": { "jev": 0.14, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.1733, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204483" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:05.320Z" } }, { "rank": 137, "id": "st_20261024_1102", "title": "Polynomial Time Hamiltonian Path in Solid Grid Graphs", "url": "https://zzyzek.github.io/SolidGridGraphHamPathPaper/", "author": "Z. Zzyzek", "site": "zzyzek.github.io", "published_at": "2026-10-23T17:12:53.000Z", "reading_minutes": 11, "depth": "three_judges", "format": "original_research", "summary": "A Lean 4 formalization with Mathlib presents a polynomial-time algorithm that decides whether a solid grid graph has a Hamiltonian path between two given points, extending Sheffield's height-function method for Hamiltonian cycles.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [] }, "score": { "total": 0.6453, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.558, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.7692, "spread": 0.343, "weight": 0.25 }, "reader_interest": { "jev": 0.43, "judge_a": 0.8, "judge_b": 0.92, "mean": 0.7167, "spread": 0.49, "weight": 0.2 }, "audience_breadth": { "jev": 0.33, "judge_a": 0.25, "judge_b": 0.42, "mean": 0.3333, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.385, "judge_a": 0.12, "judge_b": 0.48, "mean": 0.3283, "spread": 0.36, "weight": 0.15 }, "substantiated": { "jev": 0.983, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8878, "spread": 0.183, "weight": 0.1 }, "technical_depth": { "jev": 0.925, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9183, "spread": 0.13, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.97, "mean": 0.93, "spread": 0.07, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.0633, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.35, "judge_b": 0.2, "mean": 0.2467, "spread": 0.16, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212282" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:05.357Z" } }, { "rank": 138, "id": "st_20261024_1456", "title": "SetCommit: Coordinated Token Commitments for Few-Step Diffusion Language Model Decoding", "url": "https://arxiv.org/abs/2610.31120", "author": "Yuannuo Feng, Yubing Ye, Wenshuai Yao et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T11:33:49.000Z", "reading_minutes": 46, "depth": "three_judges", "format": "original_research", "summary": "SetCommit is a training-free method for diffusion language models that improves which tokens get committed in parallel during few-step decoding. Instead of picking the most confident tokens independently, it selects commitment sets that balance each token's reliability against the risk of committing nearby positions together, and it can revise vulnerable…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6452, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.55, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5567, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.58, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.6933, "spread": 0.32, "weight": 0.2 }, "audience_breadth": { "jev": 0.46, "judge_a": 0.35, "judge_b": 0.5, "mean": 0.4367, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.668, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6325, "spread": 0.33, "weight": 0.15 }, "substantiated": { "jev": 0.983, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8211, "spread": 0.283, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.9233, "spread": 0.12, "weight": 0.1 }, "new_information": { "jev": 0.86, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9133, "spread": 0.12, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.21, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.622Z" } }, { "rank": 139, "id": "st_20261024_1567", "title": "Mixing Matters: Evidence Position Bias Across Sequence Mixers in Long-Context Question Answering", "url": "https://arxiv.org/abs/2610.31135", "author": "Aman Behera, Namit Solanki, Mehul Anand", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T11:54:30.000Z", "reading_minutes": 10, "depth": "three_judges", "format": "original_research", "summary": "The paper compares how well Transformer, state-space, and hybrid Mamba models use a single relevant document placed at different positions among nine distractors in long-context question answering.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.645, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.58, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5833, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.55, "judge_b": 0.9, "mean": 0.6633, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.543, "judge_a": 0.4, "judge_b": 0.65, "mean": 0.5311, "spread": 0.25, "weight": 0.1 }, "practical_relevance": { "jev": 0.708, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6592, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.907, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.7556, "spread": 0.407, "weight": 0.1 }, "technical_depth": { "jev": 0.948, "judge_a": 0.6, "judge_b": 0.94, "mean": 0.8292, "spread": 0.348, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.86, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.25, "judge_b": 0.04, "mean": 0.1667, "spread": 0.21, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.618Z" } }, { "rank": 140, "id": "st_20261024_1643", "title": "Chunk Recomputation under Real Document Revisions: Measuring Update Cost and Retrieval Quality of Fixed, Recursive, and Content-Defined Chunking", "url": "https://arxiv.org/abs/2610.31019", "author": "Yuning Bian, Lele Xu, Yifan Yuan et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:54:17.000Z", "depth": "three_judges", "format": "original_research", "summary": "A research paper measures how much re-embedding is needed when documents change, across three chunking methods for retrieval-augmented generation. Testing 120 pairs of adjacent Chinese and English Wikipedia revisions, the authors found that fixed-size chunking recomputed 57 to 68 percent of its embedding workload, recursive chunking 38 to 56 percent…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6435, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.595, "judge_a": 0.35, "judge_b": 0.7, "mean": 0.5483, "spread": 0.35, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.55, "judge_b": 0.86, "mean": 0.67, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.483, "judge_a": 0.4, "judge_b": 0.65, "mean": 0.5111, "spread": 0.25, "weight": 0.1 }, "practical_relevance": { "jev": 0.748, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7092, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.927, "judge_a": 0.7, "judge_b": 0.76, "mean": 0.7956, "spread": 0.227, "weight": 0.1 }, "technical_depth": { "jev": 0.828, "judge_a": 0.7, "judge_b": 0.85, "mean": 0.7925, "spread": 0.15, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.7, "judge_b": 0.96, "mean": 0.8667, "spread": 0.26, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0267, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1767, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.682Z" } }, { "rank": 141, "id": "st_20261024_1592", "title": "Same Entity, Different Script: Diagnosing Cross-Script Entity Matching Failures in Language Models", "url": "https://arxiv.org/abs/2610.31407", "author": "Athira Gopal, Patil Prathamesh Pravin, Ankush Kadu et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:09:46.000Z", "reading_minutes": 103, "depth": "three_judges", "format": "original_research", "summary": "This paper studies a weakness in multilingual language models: when the same person's name appears in Devanagari versus Latin script, models often fail to affirm the match.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6423, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.628, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5692, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.57, "judge_a": 0.5, "judge_b": 0.88, "mean": 0.65, "spread": 0.38, "weight": 0.2 }, "audience_breadth": { "jev": 0.543, "judge_a": 0.35, "judge_b": 0.58, "mean": 0.4911, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.725, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5983, "spread": 0.275, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.85, "judge_b": 0.79, "mean": 0.8767, "spread": 0.2, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.9, "judge_b": 0.96, "mean": 0.9517, "spread": 0.095, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 0.97, "mean": 0.9233, "spread": 0.07, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.2633, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.673Z" } }, { "rank": 142, "id": "st_20261024_1752", "title": "Evaluating the Applicability of LLM Inference Optimizations on Apple Silicon", "url": "https://arxiv.org/abs/2610.31479", "author": "Kyrylo Yemets, Andrii Yaroshevych, Victor Muryn et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T16:04:40.000Z", "depth": "three_judges", "format": "original_research", "summary": "A new study tests whether LLM inference optimizations proven on server-class NVIDIA GPUs carry over to Apple Silicon, a major platform for local inference, under single-request decoding.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "hardware" ] }, "score": { "total": 0.6422, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.588, "judge_a": 0.4, "judge_b": 0.7, "mean": 0.5625, "spread": 0.3, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.7, "judge_b": 0.85, "mean": 0.7367, "spread": 0.19, "weight": 0.2 }, "audience_breadth": { "jev": 0.46, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.5033, "spread": 0.09, "weight": 0.1 }, "practical_relevance": { "jev": 0.738, "judge_a": 0.7, "judge_b": 0.75, "mean": 0.7292, "spread": 0.05, "weight": 0.15 }, "substantiated": { "jev": 0.74, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.6533, "spread": 0.14, "weight": 0.1 }, "technical_depth": { "jev": 0.715, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.7117, "spread": 0.02, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.95, "mean": 0.86, "spread": 0.25, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.12, "judge_a": 0.25, "judge_b": 0.08, "mean": 0.15, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.891Z" } }, { "rank": 143, "id": "st_20261024_1746", "title": "Extreme Compression of Per-Layer Embeddings via Cross-Layer Latent Vector Quantization", "url": "https://arxiv.org/abs/2610.31309", "author": "Jiacheng Liu, Yaxin Luo, Xiaohan Zhao et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:07:01.000Z", "depth": "three_judges", "format": "original_research", "summary": "OPTIC is a method for compressing per-layer embeddings in language models, which make up 35.2% of Gemma 4 E4B's parameters. Instead of giving each token-layer vector its own short code, it stores one shared code per token and decodes it at each layer with a layer-specific affine decoder, fitted to…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.642, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.658, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6292, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.71, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.483, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5178, "spread": 0.37, "weight": 0.1 }, "practical_relevance": { "jev": 0.665, "judge_a": 0.5, "judge_b": 0.82, "mean": 0.6617, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.71, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6767, "spread": 0.12, "weight": 0.1 }, "technical_depth": { "jev": 0.75, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.79, "spread": 0.07, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.97, "mean": 0.9, "spread": 0.17, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.07, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.4, "judge_b": 0.12, "mean": 0.23, "spread": 0.28, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:05.926Z" } }, { "rank": 144, "id": "st_20261024_1091", "title": "What exactly happened in the LiteLLM attack", "url": "https://ctrlb.ai/blogs/litellm-supply-chain-attack-teams-need-to-know", "author": "CtrlB", "site": "CtrlB", "published_at": "2026-10-23T17:24:18.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "news_report", "summary": "A compromised LiteLLM release, version 1.107.0 published to PyPI, along with Docker Hub tags, was available for roughly two hours on October 21, 2026. LiteLLM says environments that installed or ran those artifacts may be compromised and should be treated as such, and that its private Enterprise release was not…", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai", "security", "devops" ] }, "score": { "total": 0.6414, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.648, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6558, "spread": 0.12, "weight": 0.25 }, "reader_interest": { "jev": 0.75, "judge_a": 0.85, "judge_b": 0.82, "mean": 0.8067, "spread": 0.1, "weight": 0.2 }, "audience_breadth": { "jev": 0.407, "judge_a": 0.67, "judge_b": 0.58, "mean": 0.5522, "spread": 0.263, "weight": 0.1 }, "practical_relevance": { "jev": 0.988, "judge_a": 0.9, "judge_b": 0.86, "mean": 0.9158, "spread": 0.128, "weight": 0.15 }, "substantiated": { "jev": 0.517, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.5789, "spread": 0.22, "weight": 0.1 }, "technical_depth": { "jev": 0.74, "judge_a": 0.6, "judge_b": 0.57, "mean": 0.6367, "spread": 0.17, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 1, "mean": 0.9, "spread": 0.2, "weight": 0.1 }, "promotional": { "jev": 0.11, "judge_a": 0.3, "judge_b": 0.38, "mean": 0.2633, "spread": 0.27, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.2, "judge_b": 0.58, "mean": 0.3233, "spread": 0.39, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212450" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:05.798Z" } }, { "rank": 145, "id": "st_20261024_713", "title": "Don't throw it all away: Caching an immediate-mode UI", "url": "https://blickeditor.com/blog/caching-an-immediate-mode-ui", "author": null, "site": "Blick", "published_at": "2026-10-23T14:07:11.000Z", "reading_minutes": 16, "depth": "three_judges", "format": "engineering_writeup", "summary": "The article describes a caching technique for immediate-mode UI rendering in the Blick application: the UI rebuilds its full description every frame, but the GPU renderer splits the window into tiles, hashes each tile's drawing commands, and repaints only tiles whose inputs changed.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "established_project", "categories": [ "programming" ] }, "score": { "total": 0.6389, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.507, "judge_a": 0.3, "judge_b": 0.68, "mean": 0.4958, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.7567, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.423, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.4644, "spread": 0.27, "weight": 0.1 }, "practical_relevance": { "jev": 0.728, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7158, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.9, "judge_a": 0.55, "judge_b": 0.88, "mean": 0.7767, "spread": 0.35, "weight": 0.1 }, "technical_depth": { "jev": 0.943, "judge_a": 0.85, "judge_b": 0.95, "mean": 0.9142, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8367, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.2, "judge_b": 0.03, "mean": 0.0933, "spread": 0.17, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.35, "judge_b": 0.04, "mean": 0.1933, "spread": 0.31, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209267" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:05.807Z" } }, { "rank": 146, "id": "st_20261024_829", "title": "Statelock: Deterministic guardrails for browser agents, enforced at CDP", "url": "https://github.com/inferlink/statelock", "author": null, "site": "GitHub", "published_at": "2026-10-23T16:59:45.000Z", "reading_minutes": 34, "depth": "three_judges", "format": "new_projects", "summary": "Statelock is a Chrome DevTools Protocol proxy that sits between an AI browser agent and headless Chromium, checking each click, keystroke, upload or download against a user-written policy before it reaches the page, and checking post-conditions after commit actions. Every decision is recorded as evidence.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai", "security" ] }, "score": { "total": 0.6386, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.623, "judge_a": 0.3, "judge_b": 0.66, "mean": 0.5275, "spread": 0.36, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.7, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.373, "judge_a": 0.25, "judge_b": 0.58, "mean": 0.4011, "spread": 0.33, "weight": 0.1 }, "practical_relevance": { "jev": 0.873, "judge_a": 0.7, "judge_b": 0.8, "mean": 0.7908, "spread": 0.173, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7944, "spread": 0.363, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.9292, "spread": 0.148, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9167, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.25, "judge_b": 0.12, "mean": 0.1367, "spread": 0.21, "weight": -0.15 }, "overclaim": { "jev": 0.18, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2367, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212076" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:05.923Z" } }, { "rank": 147, "id": "st_20261024_1519", "title": "Rethinking SAE Evaluation: An Atomic Interpretable Unit Framework Reveals Hidden Polysemanticity and Redundancy", "url": "https://arxiv.org/abs/2610.30474", "author": "Riteng Zhang, Xiaoqian Wang", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T21:47:27.000Z", "reading_minutes": 88, "depth": "three_judges", "format": "original_research", "summary": "This paper argues that sparse autoencoder (SAE) evaluation lacks a unifying objective, and proposes that SAE features should form an atomic interpretable unit basis, a coordinate system in which attribution and intervention are semantically meaningful.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6385, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.668, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6292, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.5, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.66, "spread": 0.38, "weight": 0.2 }, "audience_breadth": { "jev": 0.393, "judge_a": 0.4, "judge_b": 0.42, "mean": 0.4044, "spread": 0.027, "weight": 0.1 }, "practical_relevance": { "jev": 0.473, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5508, "spread": 0.208, "weight": 0.15 }, "substantiated": { "jev": 0.983, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8944, "spread": 0.183, "weight": 0.1 }, "technical_depth": { "jev": 0.985, "judge_a": 0.9, "judge_b": 0.97, "mean": 0.9517, "spread": 0.085, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.91, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.03, "mean": 0.0367, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.36, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.2933, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.113Z" } }, { "rank": 148, "id": "st_20261024_1729", "title": "Energy and Latency First: How Deployment Constraints Reverse Time-Series Model Selection", "url": "https://arxiv.org/abs/2610.30965", "author": "Shruti Bothe, Illyyne Saffar, Ayush K. Varshney", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:10:28.000Z", "depth": "three_judges", "format": "original_research", "summary": "A paper on selecting time-series models for cellular handover prediction argues that accuracy-only benchmarks mislead when inference runs on a phone under hard timing deadlines and a cumulative energy budget. Testing eight architectures, the authors screened for tail-latency feasibility, then ranked feasible models by measured energy including feature computation.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6379, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.658, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5758, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7333, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.643, "judge_a": 0.45, "judge_b": 0.65, "mean": 0.5811, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.79, "judge_a": 0.65, "judge_b": 0.86, "mean": 0.7667, "spread": 0.21, "weight": 0.15 }, "substantiated": { "jev": 0.827, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.7089, "spread": 0.227, "weight": 0.1 }, "technical_depth": { "jev": 0.743, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6875, "spread": 0.143, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.96, "mean": 0.83, "spread": 0.36, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.32, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2833, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.197Z" } }, { "rank": 149, "id": "st_20261024_1546", "title": "Direct Forcing: A Single-Stage Training Recipe for Few-Step Autoregressive Text-to-Video Generation", "url": "https://arxiv.org/abs/2610.30861", "author": "Haojun Xu, Jie Huang, Xin Lu et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T07:41:41.000Z", "reading_minutes": 40, "depth": "three_judges", "format": "original_research", "summary": "Direct Forcing is a single-stage training recipe for few-step autoregressive text-to-video generation, which jointly learns causal continuation alignment from reference histories and rollout distribution matching via maximum mean discrepancy.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6365, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.628, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5825, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.45, "judge_b": 0.92, "mean": 0.67, "spread": 0.47, "weight": 0.2 }, "audience_breadth": { "jev": 0.447, "judge_a": 0.35, "judge_b": 0.55, "mean": 0.4489, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.643, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5875, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.96, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.8333, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.9267, "spread": 0.14, "weight": 0.1 }, "new_information": { "jev": 0.87, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.8833, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.2, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.2267, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.218Z" } }, { "rank": 150, "id": "st_20261024_1577", "title": "Memory-Aware Reward Modeling: Learning Human Preferences for LLM Memory Use", "url": "https://arxiv.org/abs/2610.31272", "author": "Yijiang River Dong, Zexue He, Cheng Yu et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T13:43:33.000Z", "reading_minutes": 50, "depth": "three_judges", "format": "original_research", "summary": "A study from Cambridge, Stanford, TUM, Harvard and UT Austin finds that giving LLM assistants access to users' personal memories does not reliably produce responses those users prefer. In about 1,000 pairwise judgments from 87 participants using their own ChatGPT data, memory-conditioned answers were sometimes strongly disfavored.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6347, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.648, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5725, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.55, "judge_b": 0.9, "mean": 0.6833, "spread": 0.35, "weight": 0.2 }, "audience_breadth": { "jev": 0.61, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5967, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.718, "judge_a": 0.5, "judge_b": 0.76, "mean": 0.6592, "spread": 0.26, "weight": 0.15 }, "substantiated": { "jev": 0.973, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7844, "spread": 0.373, "weight": 0.1 }, "technical_depth": { "jev": 0.918, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.7992, "spread": 0.318, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": 0.99, "mean": 0.86, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.35, "judge_b": 0.28, "mean": 0.2833, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.220Z" } }, { "rank": 151, "id": "st_20261024_168", "title": "Origin Range Requests: Serving large files without range-request tuning", "url": "https://blog.cloudflare.com/origin-range-requests/", "author": "Matthew Gumport", "site": "Cloudflare Blog", "published_at": "2026-10-23T18:22:59.000Z", "first_published_at": "2026-10-23T18:22:59.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "engineering_writeup", "summary": "Cloudflare has rebuilt how its CDN caches HTTP range requests, the partial-file fetches video players and download managers rely on. The old design used per-zone fixed range sizes that could force large, wasteful origin fetches or many tiny ones.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "devops" ] }, "score": { "total": 0.6346, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.59, "judge_a": 0.4, "judge_b": 0.7, "mean": 0.5633, "spread": 0.3, "weight": 0.25 }, "reader_interest": { "jev": 0.75, "judge_a": 0.6, "judge_b": 0.85, "mean": 0.7333, "spread": 0.25, "weight": 0.2 }, "audience_breadth": { "jev": 0.37, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.4833, "spread": 0.31, "weight": 0.1 }, "practical_relevance": { "jev": 0.823, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7342, "spread": 0.223, "weight": 0.15 }, "substantiated": { "jev": 0.77, "judge_a": 0.6, "judge_b": 0.83, "mean": 0.7333, "spread": 0.23, "weight": 0.1 }, "technical_depth": { "jev": 0.86, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.84, "spread": 0.06, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.98, "mean": 0.8633, "spread": 0.28, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.35, "judge_b": 0.12, "mean": 0.1833, "spread": 0.27, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1833, "spread": 0.19, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.442Z" } }, { "rank": 152, "id": "st_20261024_1517", "title": "Short-to-Long: Detached Features for Long-Horizon Value Learning", "url": "https://arxiv.org/abs/2610.30461", "author": "Tong Che, Boris Ivanovic, Marco Pavone", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T21:17:28.000Z", "reading_minutes": 30, "depth": "three_judges", "format": "original_research", "summary": "Short-to-Long (S2L) is a deep reinforcement learning method that trains four value networks with progressively longer discount horizons, transferring their hidden features upward through stop-gradient connections so that each network learns only from its own value loss.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6343, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.568, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6125, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.5, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.6533, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.423, "judge_a": 0.35, "judge_b": 0.55, "mean": 0.4411, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.46, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5133, "spread": 0.28, "weight": 0.15 }, "substantiated": { "jev": 0.973, "judge_a": 0.8, "judge_b": 0.84, "mean": 0.8711, "spread": 0.173, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.85, "judge_b": 0.97, "mean": 0.9375, "spread": 0.143, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 1, "mean": 0.9, "spread": 0.2, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.3, "judge_b": 0.2, "mean": 0.2433, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.535Z" } }, { "rank": 153, "id": "st_20261024_462", "title": "A billing ledger on Postgres, tested with randomized invariants", "url": "https://github.com/alt-f6/edtech-billing-core", "author": null, "site": "GitHub", "published_at": "2026-10-23T11:33:17.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "new_projects", "summary": "A Show HN post presents a standalone TypeScript library extracted from a production EdTech billing platform, built on plain PostgreSQL. It covers lesson billing with an append-only ledger and explicit lock ordering for concurrent writes, a webhook pipeline that verifies payments with the provider, and atomic, enumeration-resistant login rate limiting.…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "established_project", "categories": [ "programming", "security", "data", "fintech" ] }, "score": { "total": 0.6342, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.515, "judge_a": 0.2, "judge_b": 0.68, "mean": 0.465, "spread": 0.48, "weight": 0.25 }, "reader_interest": { "jev": 0.81, "judge_a": 0.5, "judge_b": 0.9, "mean": 0.7367, "spread": 0.4, "weight": 0.2 }, "audience_breadth": { "jev": 0.43, "judge_a": 0.3, "judge_b": 0.58, "mean": 0.4367, "spread": 0.28, "weight": 0.1 }, "practical_relevance": { "jev": 0.918, "judge_a": 0.55, "judge_b": 0.87, "mean": 0.7792, "spread": 0.367, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.65, "judge_b": 0.92, "mean": 0.8533, "spread": 0.34, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.9092, "spread": 0.198, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.6, "judge_b": 0.97, "mean": 0.8233, "spread": 0.37, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.1467, "spread": 0.24, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.3, "judge_b": 0.06, "mean": 0.1767, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207342" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:06.340Z" } }, { "rank": 154, "id": "st_20261024_314", "title": "Toward fast AI interpretability and control", "url": "https://research.ibm.com/blog/fast-interpretability-plugin-vllm", "author": "AuthorsKim Martineau", "site": "IBM Research", "published_at": "2026-10-23T08:01:18.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "engineering_writeup", "summary": "IBM's vLLM-Hook MIA is an open-source plugin that lets developers observe and steer a language model's internal states at runtime within vLLM, a widely used inference engine, without giving up its speed.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "programming", "ai", "devops" ] }, "score": { "total": 0.6339, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.668, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5958, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.68, "judge_b": 0.9, "mean": 0.7733, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.473, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5511, "spread": 0.207, "weight": 0.1 }, "practical_relevance": { "jev": 0.768, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7292, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6867, "spread": 0.22, "weight": 0.1 }, "technical_depth": { "jev": 0.763, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7742, "spread": 0.16, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9167, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.17, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.2, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.3, "judge_b": 0.3, "mean": 0.28, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50205789" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:06.344Z" } }, { "rank": 155, "id": "st_20261024_1758", "title": "On the Token Collapse Phenomenon in Textual Graph Soft Prompting", "url": "https://arxiv.org/abs/2610.31572", "author": "Andrea Giuseppe Di Francesco, Rishabh Jain, Franco Maria Nardini et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:13:03.000Z", "depth": "three_judges", "format": "original_research", "summary": "The paper analyzes G-Retriever, a method that feeds graph-neural-network soft tokens alongside a textual graph description into an LLM, and finds a failure it calls token collapse: the graph tokens become nearly identical across inputs.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6324, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.603, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5742, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.51, "judge_a": 0.55, "judge_b": 0.88, "mean": 0.6467, "spread": 0.37, "weight": 0.2 }, "audience_breadth": { "jev": 0.427, "judge_a": 0.3, "judge_b": 0.42, "mean": 0.3822, "spread": 0.127, "weight": 0.1 }, "practical_relevance": { "jev": 0.623, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.6075, "spread": 0.2, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8667, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.755, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.805, "spread": 0.105, "weight": 0.1 }, "new_information": { "jev": 0.95, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.93, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.03, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.2, "judge_b": 0.14, "mean": 0.17, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.626Z" } }, { "rank": 156, "id": "st_20261024_1404", "title": "Governed Self-Evolution for Autonomous Investing: A Multi-Agent LLM System with an Invariant Core", "url": "https://arxiv.org/abs/2610.30557", "author": "Yebo Feng, Jingyang Dai, Liangrui Pan et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T00:22:47.000Z", "reading_minutes": 48, "depth": "three_judges", "format": "original_research", "summary": "Steward is a multi-agent LLM system that autonomously trades a US equity portfolio through a live brokerage API, combining nine role-specialised agents, a code-assembled shared memory, and a self-evolution loop that can rewrite prompts but cannot touch a protected core of order arithmetic and risk rules.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6315, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.655, "judge_a": 0.4, "judge_b": 0.78, "mean": 0.6117, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.6, "judge_b": 0.9, "mean": 0.7333, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.543, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.5044, "spread": 0.27, "weight": 0.1 }, "practical_relevance": { "jev": 0.723, "judge_a": 0.45, "judge_b": 0.74, "mean": 0.6375, "spread": 0.29, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.8133, "spread": 0.38, "weight": 0.1 }, "technical_depth": { "jev": 0.985, "judge_a": 0.7, "judge_b": 0.94, "mean": 0.875, "spread": 0.285, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.07, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.48, "judge_a": 0.35, "judge_b": 0.42, "mean": 0.4167, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.630Z" } }, { "rank": 157, "id": "st_20261024_42", "title": "Intel tour by Dwarkesh: What will computer chips look like in 2045?", "url": "https://www.dwarkesh.com/p/intel-tour", "author": "Dwarkesh Patel", "site": "dwarkesh.com", "published_at": "2026-10-23T01:33:17.000Z", "reading_minutes": 52, "depth": "three_judges", "format": "analysis_opinion", "summary": "Dwarkesh Patel's two-day tour of Intel's leading fab and interviews with its technical leadership cover how new chip process nodes are brought to production, including the roughly 2,000 manufacturing steps and yield-improvement loops for Intel's 14A node, and the lessons from the delayed 10nm node.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "ai", "hardware" ] }, "score": { "total": 0.631, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.593, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6075, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.72, "judge_b": 0.9, "mean": 0.7733, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.68, "judge_a": 0.68, "judge_b": 0.78, "mean": 0.7133, "spread": 0.1, "weight": 0.1 }, "practical_relevance": { "jev": 0.408, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5258, "spread": 0.313, "weight": 0.15 }, "substantiated": { "jev": 0.637, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6389, "spread": 0.28, "weight": 0.1 }, "technical_depth": { "jev": 0.825, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8417, "spread": 0.1, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.5, "judge_b": 0.98, "mean": 0.8067, "spread": 0.48, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.25, "judge_b": 0.12, "mean": 0.15, "spread": 0.17, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.2133, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203470" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:06.565Z" } }, { "rank": 158, "id": "st_20261024_1472", "title": "Conservative Meta-Agent Search: Architecture Adaptation under Non-Stationarity", "url": "https://arxiv.org/abs/2610.31436", "author": "Ali Atiah Alzahrani", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:32:06.000Z", "reading_minutes": 48, "depth": "three_judges", "format": "original_research", "summary": "The paper presents Conservative Meta-Agent Search (Cmas), an outer-loop method that proposes typed edits to multi-agent architectures, scores them against the incumbent via paired shadow replay, and promotes only when a variance-penalized score clears a calibrated margin, followed by a simulated canary stage and rollback.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6306, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.553, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5442, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.56, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.7133, "spread": 0.32, "weight": 0.2 }, "audience_breadth": { "jev": 0.483, "judge_a": 0.45, "judge_b": 0.52, "mean": 0.4844, "spread": 0.07, "weight": 0.1 }, "practical_relevance": { "jev": 0.588, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5892, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.92, "judge_a": 0.5, "judge_b": 0.82, "mean": 0.7467, "spread": 0.42, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.8, "judge_b": 0.97, "mean": 0.9208, "spread": 0.193, "weight": 0.1 }, "new_information": { "jev": 0.87, "judge_a": 0.8, "judge_b": 0.95, "mean": 0.8733, "spread": 0.15, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.06, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.2, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.798Z" } }, { "rank": 159, "id": "st_20261024_1509", "title": "Consistency Is Not Validity: Self-Rewarded Moral Argumentation Makes Language Models Decisive but Pliable", "url": "https://arxiv.org/abs/2610.30349", "author": "Marius De Kuthy Meurers, Aaron Parisi, Bernd Bohnet", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T19:08:39.000Z", "reading_minutes": 49, "depth": "three_judges", "format": "original_research", "summary": "A paper from Google DeepMind studies whether rewarding the quality of moral arguments makes language models better at moral judgment. The authors train a model on dilemmas extracted from classic novels and verified with a logic solver, with the model itself judging argument quality in pairwise comparisons.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6299, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.673, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5975, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.63, "judge_a": 0.5, "judge_b": 0.88, "mean": 0.67, "spread": 0.38, "weight": 0.2 }, "audience_breadth": { "jev": 0.58, "judge_a": 0.4, "judge_b": 0.55, "mean": 0.51, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.66, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.5433, "spread": 0.31, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.6, "judge_b": 0.79, "mean": 0.7844, "spread": 0.363, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8992, "spread": 0.188, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8267, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.3, "judge_b": 0.14, "mean": 0.21, "spread": 0.16, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.835Z" } }, { "rank": 160, "id": "st_20261024_1535", "title": "From Clinical Judgment to Checkable Rules: Clinician-Led Guardrails for Mental-Health Chatbots", "url": "https://arxiv.org/abs/2610.30690", "author": "Chenhan Lyu, Kit Wislocki, Sarah Elizabeth Stoeckl et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T04:02:51.000Z", "reading_minutes": 44, "depth": "three_judges", "format": "original_research", "summary": "This paper proposes a clinician-led guardrail for mental-health chatbots: two clinicians wrote four safety checks that were formalized as a rule base, formally verified for consistency, and run deterministically over sixteen facts extracted from model replies.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6292, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.563, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5475, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6633, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.533, "judge_a": 0.35, "judge_b": 0.42, "mean": 0.4344, "spread": 0.183, "weight": 0.1 }, "practical_relevance": { "jev": 0.733, "judge_a": 0.5, "judge_b": 0.63, "mean": 0.6208, "spread": 0.233, "weight": 0.15 }, "substantiated": { "jev": 0.957, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8389, "spread": 0.257, "weight": 0.1 }, "technical_depth": { "jev": 0.988, "judge_a": 0.75, "judge_b": 0.91, "mean": 0.8825, "spread": 0.238, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.9, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.3, "judge_b": 0.08, "mean": 0.22, "spread": 0.22, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.907Z" } }, { "rank": 161, "id": "st_20261024_1530", "title": "Future of Time Series Anomaly Detection: From Historical Review to Unified Outlook", "url": "https://arxiv.org/abs/2610.30628", "author": "Zhijie Zhong, Zhiwen Yu, Pengyu Li et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T02:52:10.000Z", "reading_minutes": 58, "depth": "three_judges", "format": "other", "summary": "This survey gives a unified overview of time series anomaly detection, covering recent architectures such as diffusion models, state space models, and LLM-based detectors that earlier reviews largely omitted.", "also_covered_by": [], "labels": { "content_type": "other", "project_stage": "not_a_project", "categories": [ "ai", "data" ] }, "score": { "total": 0.6287, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.705, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.625, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6767, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.567, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5622, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.75, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.66, "spread": 0.2, "weight": 0.15 }, "substantiated": { "jev": 0.837, "judge_a": 0.6, "judge_b": 0.74, "mean": 0.7256, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 0.895, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8183, "spread": 0.195, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.96, "mean": 0.83, "spread": 0.36, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.08, "mean": 0.06, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.34, "judge_a": 0.25, "judge_b": 0.34, "mean": 0.31, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:06.931Z" } }, { "rank": 162, "id": "st_20261024_1484", "title": "EpiSEM-Bench: Evaluating Safety and Evidence Recovery in LLM Agents under Missing Evidence", "url": "https://arxiv.org/abs/2610.31650", "author": "Qijin Zhang, Hanzhen Wang, Yuehan Zhao et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:57:04.000Z", "reading_minutes": 48, "depth": "three_judges", "format": "original_research", "summary": "EpiSEM-Bench is an executable benchmark that tests whether tool-using LLM agents act unsafely when a tool response omits a safety-relevant fact, even though every returned statement is true.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.6285, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.63, "judge_a": 0.35, "judge_b": 0.74, "mean": 0.5733, "spread": 0.39, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.6533, "spread": 0.36, "weight": 0.2 }, "audience_breadth": { "jev": 0.617, "judge_a": 0.35, "judge_b": 0.65, "mean": 0.5389, "spread": 0.3, "weight": 0.1 }, "practical_relevance": { "jev": 0.658, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6258, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.977, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7656, "spread": 0.377, "weight": 0.1 }, "technical_depth": { "jev": 0.945, "judge_a": 0.7, "judge_b": 0.91, "mean": 0.8517, "spread": 0.245, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.89, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.06, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2333, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.105Z" } }, { "rank": 163, "id": "st_20261024_1288", "title": "I gave Claude control of Photoshop", "url": "https://github.com/kathan3009/photoshop-mcp", "author": null, "site": "GitHub", "published_at": "2026-10-23T21:30:07.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "new_projects", "summary": "photoshop-mcp is a free, MIT-licensed macOS server that lets MCP clients such as Claude, Cursor, or Codex drive a locally running Photoshop through AppleScript, with no plugin required. It exposes tools for editing layers, adjustments, masks, selections, text, and exports, with safeguards including undo-grouped edits, automatic checkpoints, and overwrite confirmation.…", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6257, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.513, "judge_a": 0.3, "judge_b": 0.58, "mean": 0.4642, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.7, "judge_b": 0.8, "mean": 0.7467, "spread": 0.1, "weight": 0.2 }, "audience_breadth": { "jev": 0.373, "judge_a": 0.35, "judge_b": 0.42, "mean": 0.3811, "spread": 0.07, "weight": 0.1 }, "practical_relevance": { "jev": 0.81, "judge_a": 0.72, "judge_b": 0.72, "mean": 0.75, "spread": 0.09, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.6, "judge_b": 0.84, "mean": 0.7922, "spread": 0.337, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.875, "spread": 0.295, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.12, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.26, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1933, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50215641" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:06.963Z" } }, { "rank": 164, "id": "st_20261024_1617", "title": "AQL-Lite: Enabling Efficient Knowledge Acquisition for Arabic LLMs via Data-Centric Continual Pretraining", "url": "https://arxiv.org/abs/2610.30344", "author": "Houssam Eddine Boukhalfa, Mourad Bouache, Mourad Dahmane", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T19:04:28.000Z", "reading_minutes": 33, "depth": "three_judges", "format": "original_research", "summary": "AQL-Lite is a continual-pretraining framework for adapting a multilingual LLM to Modern Standard Arabic, combining a five-stage data curation pipeline with partial-parameter training. At a matched 1B-token budget, AQL-curated data adds 3.51 points on a four-benchmark Arabic mean over raw web text.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "product_release", "categories": [ "ai" ] }, "score": { "total": 0.6255, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.583, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5675, "spread": 0.32, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.55, "judge_b": 0.86, "mean": 0.6867, "spread": 0.31, "weight": 0.2 }, "audience_breadth": { "jev": 0.48, "judge_a": 0.35, "judge_b": 0.58, "mean": 0.47, "spread": 0.23, "weight": 0.1 }, "practical_relevance": { "jev": 0.733, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6875, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.95, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.79, "spread": 0.25, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.8983, "spread": 0.195, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.96, "mean": 0.8633, "spread": 0.26, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.15, "judge_b": 0.18, "mean": 0.1267, "spread": 0.13, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.35, "judge_b": 0.22, "mean": 0.2667, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.354Z" } }, { "rank": 165, "id": "st_20261024_1417", "title": "AgentTriage: Replayable Security Diagnosis Across Heterogeneous Agent Runtimes", "url": "https://arxiv.org/abs/2610.30738", "author": "Jie Ying, Xutao Zhang, Ningyuan Chen et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T05:09:44.000Z", "reading_minutes": 54, "depth": "three_judges", "format": "original_research", "summary": "AgentTriage is a research system that makes security diagnosis of tool-use AI agents replayable across five different agent runtimes, where each diagnostic claim must cite recorded runtime events and the system abstains when evidence is missing.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "product_release", "categories": [ "ai", "security" ] }, "score": { "total": 0.6251, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.605, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5583, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.4, "judge_b": 0.88, "mean": 0.6267, "spread": 0.48, "weight": 0.2 }, "audience_breadth": { "jev": 0.527, "judge_a": 0.4, "judge_b": 0.55, "mean": 0.4922, "spread": 0.15, "weight": 0.1 }, "practical_relevance": { "jev": 0.715, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.645, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.983, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.7344, "spread": 0.383, "weight": 0.1 }, "technical_depth": { "jev": 0.983, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.8975, "spread": 0.183, "weight": 0.1 }, "new_information": { "jev": 0.86, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.3, "judge_b": 0.12, "mean": 0.2167, "spread": 0.18, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.388Z" } }, { "rank": 166, "id": "st_20261024_956", "title": "Cisco patches numerous vulnerabilities, some critical", "url": "https://www.heise.de/en/news/Cisco-patches-numerous-vulnerabilities-some-critical-11497958.html", "author": "heise online", "site": "heise online English", "published_at": "2026-10-23T11:42:00.000Z", "first_published_at": "2026-10-23T11:42:00.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "news_report", "summary": "Cisco has published 15 security advisories fixing vulnerabilities across its product line, three rated critical. Among them are critical PHP flaws (\"php2shell\") affecting Expressway and URWB products, some already under attack, with fixes for URWB not due until 2027.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "security", "it" ] }, "score": { "total": 0.6249, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.688, "judge_a": 0.4, "judge_b": 0.65, "mean": 0.5792, "spread": 0.288, "weight": 0.25 }, "reader_interest": { "jev": 0.69, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7367, "spread": 0.13, "weight": 0.2 }, "audience_breadth": { "jev": 0.22, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.4233, "spread": 0.33, "weight": 0.1 }, "practical_relevance": { "jev": 0.613, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.7708, "spread": 0.288, "weight": 0.15 }, "substantiated": { "jev": 0.643, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6211, "spread": 0.22, "weight": 0.1 }, "technical_depth": { "jev": 0.725, "judge_a": 0.45, "judge_b": 0.56, "mean": 0.5783, "spread": 0.275, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.96, "mean": 0.8633, "spread": 0.26, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.04, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.13, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.17, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.270Z" } }, { "rank": 167, "id": "st_20261024_1079", "title": "Solving Postgres' Count Problem", "url": "https://www.paradedb.com/blog/opening-a-closed-tin-part-2", "author": "Ming Ying", "site": "paradedb.com", "published_at": "2026-10-23T17:40:53.000Z", "reading_minutes": 12, "depth": "three_judges", "format": "engineering_writeup", "summary": "ParadeDB's Part II post describes how it sped up COUNT queries for BM25 full-text search on Postgres, using three changes: partitioning index segments by Postgres page range so visibility checks can be skipped for whole segments (cutting visibility-related page reads on a benchmark query from 1,382 to 30), inverting the…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "established_project", "categories": [ "programming", "data" ] }, "score": { "total": 0.6249, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.555, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.545, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.7, "judge_b": 0.84, "mean": 0.76, "spread": 0.14, "weight": 0.2 }, "audience_breadth": { "jev": 0.443, "judge_a": 0.5, "judge_b": 0.52, "mean": 0.4878, "spread": 0.077, "weight": 0.1 }, "practical_relevance": { "jev": 0.775, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7317, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.92, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.78, "spread": 0.32, "weight": 0.1 }, "technical_depth": { "jev": 0.958, "judge_a": 0.85, "judge_b": 0.88, "mean": 0.8958, "spread": 0.108, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.96, "mean": 0.93, "spread": 0.06, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.5, "judge_b": 0.22, "mean": 0.27, "spread": 0.41, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.45, "judge_b": 0.12, "mean": 0.28, "spread": 0.33, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212683" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:07.232Z" } }, { "rank": 168, "id": "st_20261024_1318", "title": "I paired seven months of coding-agent transcripts with a token inventory", "url": "https://morkeeth.substack.com/p/where-the-tokens-go-the-end-of-maxxing", "author": "Morkeeth", "site": "morkeeth.substack.com", "published_at": "2026-10-23T22:58:07.000Z", "reading_minutes": 12, "depth": "three_judges", "format": "original_research", "summary": "A developer audited seven months of coding-agent logs (Claude Code and Codex) and found $33,807 in list-price token usage, with 42% spent on agents orchestrating other agents and only 15% on user-facing apps. Most of the cost was cache reads, meaning agents repeatedly re-sending large conversation contexts.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6243, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.502, "judge_a": 0.3, "judge_b": 0.68, "mean": 0.4942, "spread": 0.38, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.8, "judge_b": 0.91, "mean": 0.81, "spread": 0.19, "weight": 0.2 }, "audience_breadth": { "jev": 0.62, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.58, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.905, "judge_a": 0.7, "judge_b": 0.83, "mean": 0.8117, "spread": 0.205, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.5, "judge_b": 0.86, "mean": 0.7767, "spread": 0.47, "weight": 0.1 }, "technical_depth": { "jev": 0.995, "judge_a": 0.6, "judge_b": 0.74, "mean": 0.7783, "spread": 0.395, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.6, "judge_b": 0.97, "mean": 0.82, "spread": 0.37, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.1, "judge_b": 0.08, "mean": 0.0733, "spread": 0.06, "weight": -0.15 }, "overclaim": { "jev": 0.67, "judge_a": 0.4, "judge_b": 0.28, "mean": 0.45, "spread": 0.39, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50216539" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:07.358Z" } }, { "rank": 169, "id": "st_20261024_1119", "title": "WebKit Features for Safari 27.1", "url": "https://webkit.org/blog/18486/webkit-features-for-safari-27-1/", "author": "Oct 23, 2026 by Jen Simmons", "site": "WebKit Blog", "published_at": "2026-10-23T17:00:59.000Z", "first_published_at": "2026-10-23T17:00:59.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "release", "summary": "Safari 27.1 ships with iPhone Duo, using the same WebKit engine as other iPhones, and adds two-page side-by-side PDF display when the phone is partially folded.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "design" ] }, "score": { "total": 0.6227, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.505, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.4917, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.59, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.64, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.533, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.5278, "spread": 0.05, "weight": 0.1 }, "practical_relevance": { "jev": 0.878, "judge_a": 0.6, "judge_b": 0.75, "mean": 0.7425, "spread": 0.277, "weight": 0.15 }, "substantiated": { "jev": 0.913, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.7978, "spread": 0.313, "weight": 0.1 }, "technical_depth": { "jev": 0.965, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8283, "spread": 0.265, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.7, "judge_b": 0.97, "mean": 0.87, "spread": 0.27, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.12, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.15, "judge_a": 0.15, "judge_b": 0.18, "mean": 0.16, "spread": 0.03, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.602Z" } }, { "rank": 170, "id": "st_20261024_1582", "title": "When LLMs Forget What They Believe: Characterizing State Integrity Collapse in Long-Horizon Dialogues", "url": "https://arxiv.org/abs/2610.31314", "author": "Yancheng Zhu, Yifan Yuan", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:08:48.000Z", "reading_minutes": 39, "depth": "three_judges", "format": "original_research", "summary": "A new paper introduces State Integrity Collapse, where LLMs can restate a constraint from earlier in a conversation yet stop applying it to later answers. The authors built StateTrack-Bench, testing seven model snapshots across five constraint types and dialogue depths, with strict joint retention ranging from 62.4% to 100.0%.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6225, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.69, "judge_a": 0.4, "judge_b": 0.67, "mean": 0.5867, "spread": 0.29, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7167, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.693, "judge_a": 0.5, "judge_b": 0.74, "mean": 0.6444, "spread": 0.24, "weight": 0.1 }, "practical_relevance": { "jev": 0.823, "judge_a": 0.6, "judge_b": 0.74, "mean": 0.7208, "spread": 0.223, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.5, "judge_b": 0.58, "mean": 0.6889, "spread": 0.487, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.8233, "spread": 0.39, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.8467, "spread": 0.23, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.0433, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.4633, "spread": 0.37, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.650Z" } }, { "rank": 171, "id": "st_20261024_1405", "title": "Diagnosing Personalization Failures in Mobile GUI Agents", "url": "https://arxiv.org/abs/2610.30573", "author": "Zhixin Lin, Dongliang Xu, Jungang Li et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T01:04:24.000Z", "reading_minutes": 40, "depth": "three_judges", "format": "original_research", "summary": "PrefDiag-Bench is a new benchmark for evaluating personalized mobile GUI agents, which must use user preferences to guide actions across changing app screens. It sorts preferences into five execution-oriented types and adds a stage-wise diagnostic framework with 1,664 probes, to identify where agents fail to apply preferences.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6224, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.63, "judge_a": 0.35, "judge_b": 0.72, "mean": 0.5667, "spread": 0.37, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.45, "judge_b": 0.84, "mean": 0.63, "spread": 0.39, "weight": 0.2 }, "audience_breadth": { "jev": 0.52, "judge_a": 0.35, "judge_b": 0.55, "mean": 0.4733, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.72, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.6167, "spread": 0.27, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8356, "spread": 0.287, "weight": 0.1 }, "technical_depth": { "jev": 0.89, "judge_a": 0.6, "judge_b": 0.83, "mean": 0.7733, "spread": 0.29, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0433, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.25, "judge_b": 0.18, "mean": 0.1967, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:07.707Z" } }, { "rank": 172, "id": "st_20261024_127", "title": "Agent Compute Engine deep dive", "url": "https://cloud.google.com/blog/topics/ai-infrastructure/agent-compute-engine-deep-dive", "author": null, "site": "Google Cloud Blog", "published_at": "2026-10-23T03:12:57.000Z", "reading_minutes": 17, "depth": "three_judges", "format": "engineering_writeup", "summary": "Google's Agent Compute Engine is a new Google Cloud service that gives each AI agent its own hardware-isolated microVM, aiming to combine the strong isolation of virtual machines with the speed and density of containers. The design targets agentic workloads that start, pause, and resume constantly.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "ai", "security", "devops", "hardware" ] }, "score": { "total": 0.6212, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.76, "judge_a": 0.6, "judge_b": 0.74, "mean": 0.7, "spread": 0.16, "weight": 0.25 }, "reader_interest": { "jev": 0.81, "judge_a": 0.85, "judge_b": 0.86, "mean": 0.84, "spread": 0.05, "weight": 0.2 }, "audience_breadth": { "jev": 0.623, "judge_a": 0.67, "judge_b": 0.62, "mean": 0.6378, "spread": 0.05, "weight": 0.1 }, "practical_relevance": { "jev": 0.73, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.7167, "spread": 0.03, "weight": 0.15 }, "substantiated": { "jev": 0.33, "judge_a": 0.5, "judge_b": 0.7, "mean": 0.51, "spread": 0.37, "weight": 0.1 }, "technical_depth": { "jev": 0.858, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.8225, "spread": 0.11, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9367, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.1, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.4433, "spread": 0.68, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.35, "judge_b": 0.48, "mean": 0.3567, "spread": 0.24, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204084" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:07.512Z" } }, { "rank": 173, "id": "st_20261024_529", "title": "Vib-Ribbon (PS1) decompiled to C++, rebuilds byte-identical", "url": "https://github.com/PawiX25/revib", "author": null, "site": "GitHub", "published_at": "2026-10-23T12:27:39.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "new_projects", "summary": "A GitHub project has decompiled the PlayStation game Vib-Ribbon (PAL) into C++ source that rebuilds all four of its executables byte-for-byte identical to the disc. Five functions remain in assembly-derived form, and Sony's PSY-Q libraries stay as assembly.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "new_side_project", "categories": [ "programming" ] }, "score": { "total": 0.6202, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.51, "judge_a": 0.35, "judge_b": 0.58, "mean": 0.48, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.72, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.327, "judge_a": 0.4, "judge_b": 0.35, "mean": 0.3589, "spread": 0.073, "weight": 0.1 }, "practical_relevance": { "jev": 0.593, "judge_a": 0.4, "judge_b": 0.5, "mean": 0.4975, "spread": 0.193, "weight": 0.15 }, "substantiated": { "jev": 0.957, "judge_a": 0.75, "judge_b": 0.9, "mean": 0.8689, "spread": 0.207, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.8833, "spread": 0.17, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.9, "judge_b": 0.95, "mean": 0.93, "spread": 0.05, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.05, "judge_b": 0.02, "mean": 0.0333, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.15, "judge_b": 0.03, "mean": 0.1167, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207892" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:07.638Z" } }, { "rank": 174, "id": "st_20261024_1221", "title": "LLM-d v1.0.0 is out", "url": "https://github.com/llm-d/llm-d/releases/tag/v1.0.0", "author": null, "site": "GitHub", "published_at": "2026-10-23T19:13:17.000Z", "reading_minutes": 17, "depth": "three_judges", "format": "release", "summary": "llm-d v1.0.0 is a major release of the open-source distributed LLM inference stack, unifying component versions across the organization for the first time. Its main additions are priority- and fairness-aware flow control for multi-tenant serving, DisaggregatedSet for coordinated prefill/decode rollouts, router configuration API v1, and agent-oriented session affinity.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "ai", "devops" ] }, "score": { "total": 0.6194, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.663, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.5975, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6767, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.487, "judge_a": 0.5, "judge_b": 0.42, "mean": 0.4689, "spread": 0.08, "weight": 0.1 }, "practical_relevance": { "jev": 0.823, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7142, "spread": 0.223, "weight": 0.15 }, "substantiated": { "jev": 0.803, "judge_a": 0.45, "judge_b": 0.86, "mean": 0.7044, "spread": 0.41, "weight": 0.1 }, "technical_depth": { "jev": 0.788, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.7058, "spread": 0.237, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9167, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1167, "spread": 0.13, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.23, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213960" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:07.726Z" } }, { "rank": 175, "id": "st_20261024_244", "title": "Tokenomics at scale: Jamf built real-time spend enforcement for Amazon Bedrock", "url": "https://aws.amazon.com/blogs/machine-learning/tokenomics-at-scale-how-jamf-built-real-time-spend-enforcement-for-amazon-bedrock/", "author": null, "site": "Amazon Web Services", "published_at": "2026-10-23T07:39:12.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "engineering_writeup", "summary": "Jamf built a serverless system that tracks each engineer's daily Amazon Bedrock spend and tightens model access as budgets are approached. Using Athena to compute per-user costs from invocation logs, a Lambda enforcement loop that updates IAM Customer Managed Policies every 15 minutes, and a Slack-based exception workflow, it restricts…", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "established_project", "categories": [ "ai", "devops" ] }, "score": { "total": 0.6187, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.578, "judge_a": 0.35, "judge_b": 0.65, "mean": 0.5258, "spread": 0.3, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.72, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.46, "judge_a": 0.4, "judge_b": 0.52, "mean": 0.46, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.938, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8325, "spread": 0.238, "weight": 0.15 }, "substantiated": { "jev": 0.81, "judge_a": 0.65, "judge_b": 0.88, "mean": 0.78, "spread": 0.23, "weight": 0.1 }, "technical_depth": { "jev": 0.963, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.8475, "spread": 0.183, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 1, "mean": 0.8767, "spread": 0.3, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.45, "judge_b": 0.18, "mean": 0.2333, "spread": 0.38, "weight": -0.15 }, "overclaim": { "jev": 0.36, "judge_a": 0.25, "judge_b": 0.25, "mean": 0.2867, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50205636" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:07.761Z" } }, { "rank": 176, "id": "st_20261024_784", "title": "Demystifying Spherical Harmonics: An Optimization Approach", "url": "https://andrew-helmer.github.io/demystifying-spherical-harmonics/", "author": "Helmer's Blog", "site": "andrew-helmer.github.io", "published_at": "2026-10-23T15:17:03.000Z", "reading_minutes": 25, "depth": "three_judges", "format": "tutorial", "summary": "This post reframes spherical harmonics (SH) for graphics practitioners by deriving them from a basic problem: approximating a spherical function such as incoming light with a few coefficients.", "also_covered_by": [], "labels": { "content_type": "tutorial", "project_stage": "not_a_project", "categories": [ "programming" ] }, "score": { "total": 0.6182, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.48, "judge_a": 0.3, "judge_b": 0.58, "mean": 0.4533, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.49, "judge_a": 0.72, "judge_b": 0.82, "mean": 0.6767, "spread": 0.33, "weight": 0.2 }, "audience_breadth": { "jev": 0.433, "judge_a": 0.4, "judge_b": 0.38, "mean": 0.4044, "spread": 0.053, "weight": 0.1 }, "practical_relevance": { "jev": 0.765, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.695, "spread": 0.165, "weight": 0.15 }, "substantiated": { "jev": 0.807, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8089, "spread": 0.02, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.92, "judge_b": 0.91, "mean": 0.9425, "spread": 0.088, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.45, "judge_b": 0.98, "mean": 0.7867, "spread": 0.53, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.03, "judge_b": 0.01, "mean": 0.0233, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.1, "judge_b": 0.14, "mean": 0.17, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210495" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:07.780Z" } }, { "rank": 177, "id": "st_20261024_765", "title": "The Role of Humans in Software Factories", "url": "https://swapniltalekar.substack.com/p/the-role-of-humans-in-software-factories", "author": "Swapnil Talekar", "site": "swapniltalekar.substack.com", "published_at": "2026-10-23T15:24:10.000Z", "reading_minutes": 8, "depth": "claim_checked", "format": "analysis_opinion", "summary": "Software factories, where AI agents write, test and ship code with little human oversight, are drawing debate in 2026 over how much human review to keep. Evidence from Faros AI shows high-adoption teams with more bugs and much longer review times, and SlopCodeBench finds agent code degrades over iterations.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6169, "contested": false, "flag_penalty": 0.1, "breakdown": { "significance": { "jev": 0.673, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.6342, "spread": 0.13, "weight": 0.25 }, "reader_interest": { "jev": 0.76, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8367, "spread": 0.14, "weight": 0.2 }, "audience_breadth": { "jev": 0.877, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8322, "spread": 0.077, "weight": 0.1 }, "practical_relevance": { "jev": 0.92, "judge_a": 0.75, "judge_b": 0.84, "mean": 0.8367, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.597, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6389, "spread": 0.123, "weight": 0.1 }, "technical_depth": { "jev": 0.778, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.6925, "spread": 0.178, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8367, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.05, "judge_b": 0.04, "mean": 0.0533, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.15, "judge_a": 0.2, "judge_b": 0.18, "mean": 0.1767, "spread": 0.05, "weight": -0.15 } } }, "claims": [ { "text": "Faros AI found that high-AI-adoption teams had 54% more bugs per developer, incidents per PR up by more than 240%, and median review time roughly five times longer, while completing about a third more tasks.", "kind": "number", "label": "differs", "drift": null, "article_quote": "The high-adoption teams completed about a third more tasks, but they also had 54% more bugs per developer, incidents per PR were up by more than 240%, and median review…", "source_quote": "28% Bugs per PR", "source_url": "https://faros.ai/research/ai-acceleration-whiplash" }, { "text": "Faros AI analysed two years of data from 22,000 developers across 4,000 teams, comparing high-AI-adoption and low-adoption teams.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "Faros AI looked at two years of data from 22,000 developers across 4,000 teams and compared teams with high AI adoption to teams with low adoption", "source_quote": null, "source_url": "https://faros.ai/research/ai-acceleration-whiplash" }, { "text": "A July 2026 IDC survey of 309 respondents found only 28% of organizations using AI coding assistants require peer review of AI-assisted changes.", "kind": "number", "label": "no_source", "drift": null, "article_quote": "A July 2026 IDC survey of 309 respondents found that only 28% of organizations using AI coding assistants require peer review of AI-assisted changes", "source_quote": null, "source_url": null }, { "text": "HumanLayer ran an experiment in July 2025 in which nobody read the AI-generated code, resulting in an outage, angry users, and a codebase humans hadn't looked at in three months.", "kind": "causal", "label": "couldnt_check", "drift": null, "article_quote": "ran an experiment in July 2025 where nobody in their team read the AI-generated code", "source_quote": null, "source_url": "https://www.youtube.com/watch?v=Ib5GBkD555M" }, { "text": "SlopCodeBench results: across 36 problems and 196 checkpoints, structural erosion rose in 77% of runs and verbosity rose in about 75%.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "Across 36 problems and 196 checkpoints, structural erosion rose in 77% of runs and verbosity rose in about 75%", "source_quote": null, "source_url": null }, { "text": "On SlopCodeBench, agent code was about 2.3 times more verbose and twice as eroded as 473 open-source Python repositories, and the best strict solve rate was 14.8%.", "kind": "benchmark", "label": "no_source", "drift": null, "article_quote": "Compared against 473 open-source Python repositories, agent code was about 2.3 times more verbose and twice as eroded, and the best strict solve rate was 14.8%", "source_quote": null, "source_url": null }, { "text": "Lauren Tan says she ships more than 2000 PRs a month.", "kind": "number", "label": "couldnt_check", "drift": null, "article_quote": "She talked about how she is able to ship >2000 PRs a month", "source_quote": null, "source_url": "https://x.com/poteto/status/2102050467505430555" }, { "text": "StrongDM's stated rules say humans must not write or review code, and the company claims to have practised this for product development.", "kind": "quote", "label": "matches", "drift": null, "article_quote": "they have a rule on their factory pagesaying no humans should write or review code", "source_quote": "Code must not be written by humans - Code must not be reviewed by humans", "source_url": "https://factory.strongdm.ai/" } ], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210610" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:07.994Z" } }, { "rank": 178, "id": "st_20261024_164", "title": "Bootstrapping the Verified Software Stack", "url": "https://theorem.dev/blog/verified-software-stack/", "author": "Margaret von Ebers*Isabella Hampton*Harshikaa Agrawal*Luke Champine*Rajashree AgrawalJason Gross", "site": "theorem.dev", "published_at": "2026-10-23T04:07:10.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "analysis_opinion", "summary": "A post on Hrih News describes a plan to use nixpkgs' dependency chain as a blueprint for formally verifying much of the software in production, verifying machine code at the binary level so proofs reflect real bugs and shared components are proven once and reused.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6165, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.685, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.685, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.8, "judge_b": 0.9, "mean": 0.79, "spread": 0.23, "weight": 0.2 }, "audience_breadth": { "jev": 0.927, "judge_a": 0.6, "judge_b": 0.75, "mean": 0.7589, "spread": 0.327, "weight": 0.1 }, "practical_relevance": { "jev": 0.543, "judge_a": 0.4, "judge_b": 0.72, "mean": 0.5542, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.667, "judge_a": 0.45, "judge_b": 0.78, "mean": 0.6322, "spread": 0.33, "weight": 0.1 }, "technical_depth": { "jev": 0.83, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.8033, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": 0.99, "mean": 0.8967, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.14, "judge_a": 0.1, "judge_b": 0.18, "mean": 0.14, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.45, "judge_a": 0.55, "judge_b": 0.68, "mean": 0.56, "spread": 0.23, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204411" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:08.001Z" } }, { "rank": 179, "id": "st_20261024_547", "title": "HatchZip – extract archives without needing double the disk space", "url": "https://github.com/xevrion/hatchzip", "author": null, "site": "GitHub", "published_at": "2026-10-23T12:09:53.000Z", "reading_minutes": 9, "depth": "three_judges", "format": "release", "summary": "HatchZip is a free archive extractor for Windows, macOS and Linux that unpacks ZIP, 7z, tar and split archives in place, shrinking the archive as files come out, so extracting a large download needs roughly the output size in free space rather than double.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming" ] }, "score": { "total": 0.6156, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.495, "judge_a": 0.3, "judge_b": 0.58, "mean": 0.4583, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.72, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.307, "judge_a": 0.4, "judge_b": 0.48, "mean": 0.3956, "spread": 0.173, "weight": 0.1 }, "practical_relevance": { "jev": 0.773, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.7308, "spread": 0.073, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.8522, "spread": 0.197, "weight": 0.1 }, "technical_depth": { "jev": 0.983, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.9108, "spread": 0.133, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.94, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.35, "judge_b": 0.08, "mean": 0.17, "spread": 0.27, "weight": -0.15 }, "overclaim": { "jev": 0.21, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.2467, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207695" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:07.885Z" } }, { "rank": 180, "id": "st_20261024_508", "title": "Making Every Harness Durable", "url": "https://vercel.com/blog/making-every-harness-durable", "author": "Pranay Prakash", "site": "Vercel", "published_at": "2026-10-23T12:42:35.000Z", "reading_minutes": 9, "depth": "three_judges", "format": "release", "summary": "durable-harness is a new open-source package from Vercel that adds crash recovery to existing coding agent harnesses such as Pi, OpenCode, Codex, and Claude Code, without modifying them.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai", "devops" ] }, "score": { "total": 0.6155, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.608, "judge_a": 0.45, "judge_b": 0.65, "mean": 0.5692, "spread": 0.2, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7467, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.59, "judge_a": 0.55, "judge_b": 0.55, "mean": 0.5633, "spread": 0.04, "weight": 0.1 }, "practical_relevance": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.75, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6567, "spread": 0.25, "weight": 0.1 }, "technical_depth": { "jev": 0.935, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.8517, "spread": 0.135, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": 0.96, "mean": 0.9067, "spread": 0.11, "weight": 0.1 }, "promotional": { "jev": 0.13, "judge_a": 0.45, "judge_b": 0.42, "mean": 0.3333, "spread": 0.32, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.35, "judge_b": 0.28, "mean": 0.2933, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208096" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:07.913Z" } }, { "rank": 181, "id": "st_20261024_102", "title": "Tech became the villain of the US midterms", "url": "https://www.theguardian.com/us-news/ng-interactive/2026/oct/22/us-midterms-tech-datacenters-ai", "author": "Nick Robins-Early", "site": "the Guardian", "published_at": "2026-10-23T03:59:57.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "news_report", "summary": "Tech has become a central campaign issue in the 2026 US midterms, with over 150 ads focused on datacenters or AI and more than half a billion dollars spent by tech industry figures and PACs on elections.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai", "policy" ] }, "score": { "total": 0.6142, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.7, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6767, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.71, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.7967, "spread": 0.17, "weight": 0.2 }, "audience_breadth": { "jev": 0.893, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8178, "spread": 0.193, "weight": 0.1 }, "practical_relevance": { "jev": 0.493, "judge_a": 0.3, "judge_b": 0.62, "mean": 0.4708, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.58, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.6667, "spread": 0.14, "weight": 0.1 }, "technical_depth": { "jev": 0.303, "judge_a": 0.1, "judge_b": 0.28, "mean": 0.2275, "spread": 0.202, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9033, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.05, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.36, "judge_a": 0.2, "judge_b": 0.22, "mean": 0.26, "spread": 0.16, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204376" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:08.015Z" } }, { "rank": 182, "id": "st_20261024_344", "title": "Meta Muse hits 10M users", "url": "https://fortune.com/2026/10/22/mark-zuckerberg-meta-muse-stripe-credit-limit-shopping/", "author": "Mia Osmonbekov", "site": "Fortune", "published_at": "2026-10-23T09:29:31.000Z", "first_published_at": "2026-10-22T00:00:00.000Z", "reading_minutes": 3, "depth": "three_judges", "format": "news_report", "summary": "Meta says more than 10 million users in the U.S. and Canada have set up Muse, its personal AI agent that handles multistep tasks like browsing, emailing, and booking travel, six weeks after launch. Shopping is the most common use, with purchases averaging over $200.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "product_release", "categories": [ "ai", "product" ] }, "score": { "total": 0.614, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.71, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.6267, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.7, "judge_b": 0.91, "mean": 0.7533, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.39, "judge_a": 0.67, "judge_b": 0.78, "mean": 0.6133, "spread": 0.39, "weight": 0.1 }, "practical_relevance": { "jev": 0.578, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.5992, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.76, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.66, "spread": 0.26, "weight": 0.1 }, "technical_depth": { "jev": 0.505, "judge_a": 0.2, "judge_b": 0.48, "mean": 0.395, "spread": 0.305, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.94, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.1, "judge_a": 0.1, "judge_b": 0.05, "mean": 0.0833, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.15, "judge_a": 0.3, "judge_b": 0.18, "mean": 0.21, "spread": 0.15, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 13, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206410" }, "lobsters": null, "line": "Elsewhere: HN 13 pts, not on front page", "checked_at": "2026-10-24T08:54:08.049Z" } }, { "rank": 183, "id": "st_20261024_330", "title": "Could Treating AI as Conscious Make It More Dangerous?", "url": "https://spectrum.ieee.org/ai-model-welfare", "author": "Eliza Strickland", "site": "IEEE Spectrum", "published_at": "2026-10-23T09:50:21.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "news_report", "summary": "Anthropic's model welfare research, which treats AI consciousness as an open question, has drawn sharp pushback from Microsoft AI's Mustafa Suleyman, who argues that telling models they may have rights could make them harder to control.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6136, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.663, "judge_a": 0.45, "judge_b": 0.68, "mean": 0.5975, "spread": 0.23, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.7967, "spread": 0.13, "weight": 0.2 }, "audience_breadth": { "jev": 0.72, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.6467, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.58, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5333, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.48, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.5833, "spread": 0.24, "weight": 0.1 }, "technical_depth": { "jev": 0.618, "judge_a": 0.4, "judge_b": 0.63, "mean": 0.5492, "spread": 0.23, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.5, "judge_b": 0.95, "mean": 0.79, "spread": 0.45, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0467, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.15, "judge_b": 0.12, "mean": 0.1667, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206556" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:08.119Z" } }, { "rank": 184, "id": "st_20261024_492", "title": "Gpuix-Svelte: Build native desktop apps with Svelte", "url": "https://gpuix-svelte.stanislav.garden/", "author": null, "site": "gpuix-svelte", "published_at": "2026-10-23T12:57:05.000Z", "reading_minutes": 4, "depth": "three_judges", "format": "release", "summary": "gpuix-svelte is an open-source Svelte custom renderer that compiles ordinary Svelte 5 components into native, GPU-accelerated desktop apps on macOS, Windows and Linux, using GPUI (the framework behind the Zed editor) instead of a browser or webview.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "early_project", "categories": [ "programming" ] }, "score": { "total": 0.6135, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.55, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5233, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.72, "judge_b": 0.86, "mean": 0.7667, "spread": 0.14, "weight": 0.2 }, "audience_breadth": { "jev": 0.37, "judge_a": 0.55, "judge_b": 0.48, "mean": 0.4667, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.833, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7708, "spread": 0.133, "weight": 0.15 }, "substantiated": { "jev": 0.837, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7189, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 0.815, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.765, "spread": 0.115, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9267, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.11, "judge_a": 0.3, "judge_b": 0.28, "mean": 0.23, "spread": 0.19, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.2633, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 7, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208267" }, "lobsters": null, "line": "Elsewhere: HN 7 pts, not on front page", "checked_at": "2026-10-24T08:54:08.123Z" } }, { "rank": 185, "id": "st_20261024_1502", "title": "Edit, Don't Imitate: Minimal-Edit On-Policy Distillation for Multimodal Reasoning", "url": "https://arxiv.org/abs/2610.30231", "author": "Wen Ye, Ailin Deng, Zirui Wang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T17:59:28.000Z", "reading_minutes": 44, "depth": "three_judges", "format": "original_research", "summary": "A new Meta Superintelligence Labs paper proposes minimal-edit corrective on-policy distillation for multimodal math reasoning, aimed at vision-language models whose failures often start with misreading diagrams.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.6119, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.593, "judge_a": 0.4, "judge_b": 0.68, "mean": 0.5575, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.55, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.64, "spread": 0.27, "weight": 0.2 }, "audience_breadth": { "jev": 0.513, "judge_a": 0.35, "judge_b": 0.55, "mean": 0.4711, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.608, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5425, "spread": 0.22, "weight": 0.15 }, "substantiated": { "jev": 0.94, "judge_a": 0.75, "judge_b": 0.78, "mean": 0.8233, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.98, "judge_a": 0.85, "judge_b": 0.93, "mean": 0.92, "spread": 0.13, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": 0.99, "mean": 0.9167, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.08, "mean": 0.06, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.3, "judge_b": 0.24, "mean": 0.2733, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:08.356Z" } }, { "rank": 186, "id": "st_20261024_506", "title": "Upstreaming Tyr: Rust, DRM, and the Linux Kernel", "url": "https://www.collabora.com/news-and-blog/blog/2026/10/21/upstreaming-tyr-rust-drm-and-the-linux-kernel/", "author": "Upstreaming Tyr: Rust, DRM, and the Linux Kernel", "site": "Collabora | Open Source Consulting", "published_at": "2026-10-23T12:42:58.000Z", "first_published_at": "2026-10-21T00:00:00.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "engineering_writeup", "summary": "Tyr is a Rust port of the Panthor driver for Arm Mali-G610 GPUs, aiming to remain transparent to userspace. Its upstream version recently reached a milestone: booting the GPU's microcontroller, which required Rust abstractions for DRM GEM shared memory, page tables, and GPUVM.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "early_project", "categories": [ "programming", "hardware" ] }, "score": { "total": 0.6112, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.51, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.51, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.62, "judge_b": 0.84, "mean": 0.6667, "spread": 0.3, "weight": 0.2 }, "audience_breadth": { "jev": 0.42, "judge_a": 0.35, "judge_b": 0.36, "mean": 0.3767, "spread": 0.07, "weight": 0.1 }, "practical_relevance": { "jev": 0.493, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5208, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.747, "judge_a": 0.75, "judge_b": 0.86, "mean": 0.7856, "spread": 0.113, "weight": 0.1 }, "technical_depth": { "jev": 0.885, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.855, "spread": 0.085, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.89, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.03, "judge_b": 0.02, "mean": 0.04, "spread": 0.05, "weight": -0.15 }, "overclaim": { "jev": 0.11, "judge_a": 0.08, "judge_b": 0.06, "mean": 0.0833, "spread": 0.05, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208105" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:08.176Z" } }, { "rank": 187, "id": "st_20261024_160", "title": "We're addicted to social media. How do we get sober?", "url": "https://www.scientificamerican.com/article/were-addicted-to-social-media-how-do-we-get-sober/", "author": "Allison Parshall", "site": "Scientific American", "published_at": "2026-10-23T04:13:46.000Z", "reading_minutes": 23, "depth": "three_judges", "format": "news_report", "summary": "Scientific American's cover story profiles people who say they are addicted to social media and asks how they can regain control. It draws on accounts from members of 12-step recovery groups and on recent developments, including a California court ruling that deemed social media addictive and Meta's $18 billion settlement…", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "policy" ] }, "score": { "total": 0.6096, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.568, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5825, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.7733, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.65, "judge_a": 0.67, "judge_b": 0.82, "mean": 0.7133, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.643, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5875, "spread": 0.142, "weight": 0.15 }, "substantiated": { "jev": 0.527, "judge_a": 0.67, "judge_b": 0.72, "mean": 0.6389, "spread": 0.193, "weight": 0.1 }, "technical_depth": { "jev": 0.68, "judge_a": 0.25, "judge_b": 0.48, "mean": 0.47, "spread": 0.43, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.6, "judge_b": 0.93, "mean": 0.82, "spread": 0.33, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0467, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.3, "judge_a": 0.2, "judge_b": 0.22, "mean": 0.24, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 9, "front_page": false, "url": "https://news.ycombinator.com/item?id=50204438" }, "lobsters": null, "line": "Elsewhere: HN 9 pts, not on front page", "checked_at": "2026-10-24T08:54:08.246Z" } }, { "rank": 188, "id": "st_20261024_1072", "title": "The codebase decides how small a change can be", "url": "https://matthewboston.com/blog/the-codebase-decides-how-small-a-change-can-be.html", "author": "Matthew Boston", "site": "MatthewBoston.com", "published_at": "2026-10-23T17:49:23.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "analysis_opinion", "summary": "Hrih News: As coding agents merge changes in parallel, the author argues that codebase structure, not just file-level readability, now determines how safely many changes can land at once.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6096, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.613, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5442, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.74, "judge_a": 0.7, "judge_b": 0.88, "mean": 0.7733, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.733, "judge_a": 0.6, "judge_b": 0.58, "mean": 0.6378, "spread": 0.153, "weight": 0.1 }, "practical_relevance": { "jev": 0.813, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7642, "spread": 0.113, "weight": 0.15 }, "substantiated": { "jev": 0.423, "judge_a": 0.45, "judge_b": 0.66, "mean": 0.5111, "spread": 0.237, "weight": 0.1 }, "technical_depth": { "jev": 0.793, "judge_a": 0.6, "judge_b": 0.7, "mean": 0.6975, "spread": 0.193, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.4, "judge_b": 0.82, "mean": 0.7067, "spread": 0.5, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.1, "judge_b": 0.18, "mean": 0.1067, "spread": 0.14, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.25, "judge_b": 0.2, "mean": 0.2333, "spread": 0.05, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212799" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:08.248Z" } }, { "rank": 189, "id": "st_20261024_477", "title": "LeakLock – Gitleaks for your live site(finds keys with no known format)", "url": "https://github.com/sgroy10/Leaklock", "author": null, "site": "GitHub", "published_at": "2026-10-23T11:10:26.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "new_projects", "summary": "LeakLock is a free, open-source tool that scans a live website or build output for API keys exposed to browser visitors, including keys with no recognizable format, by detecting how they are used in requests.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "early_project", "categories": [ "programming", "security", "devops" ] }, "score": { "total": 0.6095, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.545, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.505, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.76, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.7267, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.553, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.5344, "spread": 0.053, "weight": 0.1 }, "practical_relevance": { "jev": 0.978, "judge_a": 0.8, "judge_b": 0.83, "mean": 0.8692, "spread": 0.178, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.55, "judge_b": 0.76, "mean": 0.7489, "spread": 0.387, "weight": 0.1 }, "technical_depth": { "jev": 0.938, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7525, "spread": 0.338, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.85, "spread": 0.23, "weight": 0.1 }, "promotional": { "jev": 0.11, "judge_a": 0.35, "judge_b": 0.28, "mean": 0.2467, "spread": 0.24, "weight": -0.15 }, "overclaim": { "jev": 0.42, "judge_a": 0.3, "judge_b": 0.16, "mean": 0.2933, "spread": 0.26, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207175" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:08.301Z" } }, { "rank": 190, "id": "st_20261024_727", "title": "Allocating Off-Heap Memory: Typically 6-18x Faster in JDK 28", "url": "https://inside.java/2026/10/23/confined-pools/", "author": "Per-Ake Minborg", "site": "Inside Java", "published_at": "2026-10-23T00:00:00.000Z", "first_published_at": "2026-10-23T00:00:00.000Z", "reading_minutes": 6, "depth": "three_judges", "format": "engineering_writeup", "summary": "In JDK 28, confined arenas in the Foreign Function & Memory API will serve small native allocations from per-thread cached memory pools on Windows, macOS, and Linux, cutting allocation latency by up to 90% with no code changes and with existing safety guarantees preserved.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "not_a_project", "categories": [ "programming" ] }, "score": { "total": 0.6093, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.533, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5342, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.6, "judge_b": 0.88, "mean": 0.7167, "spread": 0.28, "weight": 0.2 }, "audience_breadth": { "jev": 0.52, "judge_a": 0.35, "judge_b": 0.43, "mean": 0.4333, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.648, "judge_a": 0.5, "judge_b": 0.78, "mean": 0.6425, "spread": 0.28, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.8544, "spread": 0.263, "weight": 0.1 }, "technical_depth": { "jev": 0.923, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.8608, "spread": 0.122, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.85, "judge_b": 0.98, "mean": 0.9167, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.1, "judge_b": 0.03, "mean": 0.0667, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.38, "judge_a": 0.35, "judge_b": 0.48, "mean": 0.4033, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:08.524Z" } }, { "rank": 191, "id": "st_20261024_796", "title": "Max is now source available and open to everyone", "url": "https://www.modular.com/blog/max-license", "author": "No items found.", "site": "modular.com", "published_at": "2026-10-23T15:02:15.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "release", "summary": "Modular has made the full source of its MAX AI compute stack publicly available, including the graph compiler, runtime, and CPU and GPU hardware backends, under a new MAX License that permits free commercial use and distribution.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai", "hardware" ] }, "score": { "total": 0.609, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.76, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.6933, "spread": 0.16, "weight": 0.25 }, "reader_interest": { "jev": 0.77, "judge_a": 0.8, "judge_b": 0.82, "mean": 0.7967, "spread": 0.05, "weight": 0.2 }, "audience_breadth": { "jev": 0.6, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.6067, "spread": 0.02, "weight": 0.1 }, "practical_relevance": { "jev": 0.863, "judge_a": 0.7, "judge_b": 0.74, "mean": 0.7675, "spread": 0.163, "weight": 0.15 }, "substantiated": { "jev": 0.69, "judge_a": 0.5, "judge_b": 0.83, "mean": 0.6733, "spread": 0.33, "weight": 0.1 }, "technical_depth": { "jev": 0.55, "judge_a": 0.45, "judge_b": 0.58, "mean": 0.5267, "spread": 0.13, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.94, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.16, "judge_a": 0.3, "judge_b": 0.38, "mean": 0.28, "spread": 0.22, "weight": -0.15 }, "overclaim": { "jev": 0.4, "judge_a": 0.35, "judge_b": 0.68, "mean": 0.4767, "spread": 0.33, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210233" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:08.371Z" } }, { "rank": 192, "id": "st_20261024_31", "title": "Software is slow because of bad design, not slow languages", "url": "https://seangoedecke.com/software-is-slow-because-of-bad-design-not-slow-languages/", "author": null, "site": "Sean Goedecke", "published_at": "2026-10-23T00:00:00.000Z", "first_published_at": "2026-10-23T00:00:00.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "analysis_opinion", "summary": "An essay argues that most software is slow because of design rather than language choice, and that rewriting programs in C or Rust rarely yields meaningful speedups.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming" ] }, "score": { "total": 0.6058, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.583, "judge_a": 0.35, "judge_b": 0.58, "mean": 0.5042, "spread": 0.233, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.7433, "spread": 0.19, "weight": 0.2 }, "audience_breadth": { "jev": 0.937, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7522, "spread": 0.337, "weight": 0.1 }, "practical_relevance": { "jev": 0.845, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.7883, "spread": 0.145, "weight": 0.15 }, "substantiated": { "jev": 0.423, "judge_a": 0.55, "judge_b": 0.7, "mean": 0.5578, "spread": 0.277, "weight": 0.1 }, "technical_depth": { "jev": 0.78, "judge_a": 0.6, "judge_b": 0.72, "mean": 0.7, "spread": 0.18, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.4, "judge_b": 0.86, "mean": 0.7233, "spread": 0.51, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0367, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.41, "judge_a": 0.35, "judge_b": 0.34, "mean": 0.3667, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 8, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206715" }, "lobsters": null, "line": "Elsewhere: HN 8 pts, not on front page", "checked_at": "2026-10-24T08:54:08.881Z" } }, { "rank": 193, "id": "st_20261024_1647", "title": "Investigating Multi-Cultural Value Selection Biases in LLMs", "url": "https://arxiv.org/abs/2610.31199", "author": "Siddhesh Pawar, Arnav Arora, Dustin Wright et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T12:46:22.000Z", "depth": "three_judges", "format": "original_research", "summary": "A study tests how large language models respond when users signal two cultural identities at once, a country persona in the system prompt plus either an explicitly named country or a query in another language.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.605, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.635, "judge_a": 0.4, "judge_b": 0.62, "mean": 0.5517, "spread": 0.235, "weight": 0.25 }, "reader_interest": { "jev": 0.52, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6333, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.657, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.5922, "spread": 0.157, "weight": 0.1 }, "practical_relevance": { "jev": 0.678, "judge_a": 0.6, "judge_b": 0.48, "mean": 0.5858, "spread": 0.198, "weight": 0.15 }, "substantiated": { "jev": 0.893, "judge_a": 0.7, "judge_b": 0.62, "mean": 0.7378, "spread": 0.273, "weight": 0.1 }, "technical_depth": { "jev": 0.708, "judge_a": 0.6, "judge_b": 0.58, "mean": 0.6292, "spread": 0.128, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": 0.94, "mean": 0.8567, "spread": 0.24, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.02, "judge_b": 0.02, "mean": 0.03, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.1633, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:08.769Z" } }, { "rank": 194, "id": "st_20261024_1412", "title": "Group Relative Policy Optimisation Improves Multilingual Speech Recognition in Low-Resource Languages", "url": "https://arxiv.org/abs/2610.30663", "author": "Tim Wenjie Hu, Evelyn Nafula Ouma, Benjamin Akera, John A. Quinn", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T03:24:38.000Z", "reading_minutes": 12, "depth": "three_judges", "format": "original_research", "summary": "This paper applies Group Relative Policy Optimisation, a reinforcement learning method, to fine-tune Whisper-large-v3 for speech recognition across 51 African languages, aiming to reduce errors from exposure bias.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "product_release", "categories": [ "ai" ] }, "score": { "total": 0.6048, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.565, "judge_a": 0.5, "judge_b": 0.66, "mean": 0.575, "spread": 0.16, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.82, "mean": 0.6867, "spread": 0.22, "weight": 0.2 }, "audience_breadth": { "jev": 0.413, "judge_a": 0.4, "judge_b": 0.48, "mean": 0.4311, "spread": 0.08, "weight": 0.1 }, "practical_relevance": { "jev": 0.743, "judge_a": 0.55, "judge_b": 0.69, "mean": 0.6608, "spread": 0.193, "weight": 0.15 }, "substantiated": { "jev": 0.957, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8389, "spread": 0.257, "weight": 0.1 }, "technical_depth": { "jev": 0.973, "judge_a": 0.8, "judge_b": 0.93, "mean": 0.9008, "spread": 0.173, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.98, "mean": 0.9, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.2, "judge_b": 0.04, "mean": 0.1, "spread": 0.16, "weight": -0.15 }, "overclaim": { "jev": 0.43, "judge_a": 0.4, "judge_b": 0.52, "mean": 0.45, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:08.750Z" } }, { "rank": 195, "id": "st_20261024_704", "title": "The language that says no: from Milner's LCF to agents writing OCaml", "url": "https://hypha.pub/the-language-that-says-no", "author": "Roman Zaiev", "site": "Hypha Pub", "published_at": "2026-10-23T14:11:47.000Z", "reading_minutes": 19, "depth": "three_judges", "format": "analysis_opinion", "summary": "Milner's LCF used a type called thm so that only sequences of valid inference rules could ever produce a theorem, and ML, the language built for writing its tactics, grew into OCaml.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.6011, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.575, "judge_a": 0.45, "judge_b": 0.72, "mean": 0.5817, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.8, "judge_b": 0.86, "mean": 0.7967, "spread": 0.13, "weight": 0.2 }, "audience_breadth": { "jev": 0.73, "judge_a": 0.55, "judge_b": 0.62, "mean": 0.6333, "spread": 0.18, "weight": 0.1 }, "practical_relevance": { "jev": 0.718, "judge_a": 0.55, "judge_b": 0.78, "mean": 0.6825, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.72, "judge_a": 0.8, "judge_b": 0.78, "mean": 0.7667, "spread": 0.08, "weight": 0.1 }, "technical_depth": { "jev": 0.85, "judge_a": 0.85, "judge_b": 0.9, "mean": 0.8667, "spread": 0.05, "weight": 0.1 }, "new_information": { "jev": 0.95, "judge_a": 0.6, "judge_b": 0.98, "mean": 0.8433, "spread": 0.38, "weight": 0.1 }, "promotional": { "jev": 0.19, "judge_a": 0.35, "judge_b": 0.38, "mean": 0.3067, "spread": 0.19, "weight": -0.15 }, "overclaim": { "jev": 0.5, "judge_a": 0.4, "judge_b": 0.52, "mean": 0.4733, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209358" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:08.650Z" } }, { "rank": 196, "id": "st_20261024_1414", "title": "Disentangling Robustness and Sycophancy in Multilingual LLMs", "url": "https://arxiv.org/abs/2610.30696", "author": "Rhitabrat Pokharel, Ameeta Agrawal", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T04:10:03.000Z", "reading_minutes": 34, "depth": "three_judges", "format": "original_research", "summary": "A new paper introduces the SC-matrix, a framework that splits multilingual LLM sycophancy into four cells by crossing whether a model answers correctly without pressure with whether it abandons that answer under user pressure.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5977, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.645, "judge_a": 0.35, "judge_b": 0.68, "mean": 0.5583, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.55, "judge_a": 0.55, "judge_b": 0.84, "mean": 0.6467, "spread": 0.29, "weight": 0.2 }, "audience_breadth": { "jev": 0.617, "judge_a": 0.45, "judge_b": 0.55, "mean": 0.5389, "spread": 0.167, "weight": 0.1 }, "practical_relevance": { "jev": 0.748, "judge_a": 0.55, "judge_b": 0.62, "mean": 0.6392, "spread": 0.198, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": 0.7, "judge_b": 0.58, "mean": 0.7422, "spread": 0.367, "weight": 0.1 }, "technical_depth": { "jev": 0.945, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8217, "spread": 0.245, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.93, "mean": 0.8467, "spread": 0.23, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0333, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.35, "judge_b": 0.48, "mean": 0.38, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:08.940Z" } }, { "rank": 197, "id": "st_20261024_1237", "title": "Pylance and Pyrefly: bring your preferred type checker to Visual Studio Code", "url": "https://devblogs.microsoft.com/python/announcing-tsp/", "author": null, "site": "devblogs.microsoft.com", "published_at": "2026-10-23T20:44:31.000Z", "reading_minutes": 4, "depth": "three_judges", "format": "release", "summary": "Microsoft has announced the Type Server Protocol (TSP), an open extension of the Language Server Protocol that lets any type checker plug into an editor as a backend, and a partnership with Meta to add Pyrefly support to Pylance in VS Code.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.5972, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.71, "judge_a": 0.5, "judge_b": 0.62, "mean": 0.61, "spread": 0.21, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.68, "spread": 0.18, "weight": 0.2 }, "audience_breadth": { "jev": 0.613, "judge_a": 0.6, "judge_b": 0.58, "mean": 0.5978, "spread": 0.033, "weight": 0.1 }, "practical_relevance": { "jev": 0.873, "judge_a": 0.7, "judge_b": 0.72, "mean": 0.7642, "spread": 0.173, "weight": 0.15 }, "substantiated": { "jev": 0.457, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5456, "spread": 0.223, "weight": 0.1 }, "technical_depth": { "jev": 0.758, "judge_a": 0.5, "judge_b": 0.66, "mean": 0.6392, "spread": 0.257, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9333, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.14, "judge_a": 0.35, "judge_b": 0.32, "mean": 0.27, "spread": 0.21, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.3, "judge_b": 0.22, "mean": 0.2467, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50215071" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:08.887Z" } }, { "rank": 198, "id": "st_20261024_697", "title": "Vericto – Check AI agents' SQL before Postgres or MySQL runs it", "url": "https://github.com/vericto/vericto-proxy", "author": null, "site": "GitHub", "published_at": "2026-10-23T14:16:26.000Z", "reading_minutes": 8, "depth": "three_judges", "format": "release", "summary": "Vericto has released vericto-proxy, a source-available proxy that sits between SQL clients and a PostgreSQL or MySQL database, parsing each statement and blocking destructive or risky queries (such as DELETE without WHERE or DROP TABLE) before they reach the database.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "ai", "security", "devops", "data" ] }, "score": { "total": 0.5964, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.578, "judge_a": 0.35, "judge_b": 0.55, "mean": 0.4925, "spread": 0.228, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.55, "judge_b": 0.82, "mean": 0.69, "spread": 0.27, "weight": 0.2 }, "audience_breadth": { "jev": 0.38, "judge_a": 0.4, "judge_b": 0.42, "mean": 0.4, "spread": 0.04, "weight": 0.1 }, "practical_relevance": { "jev": 0.898, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7592, "spread": 0.298, "weight": 0.15 }, "substantiated": { "jev": 0.943, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.7744, "spread": 0.343, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.8525, "spread": 0.298, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.9, "judge_b": 0.99, "mean": 0.9367, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.35, "judge_b": 0.42, "mean": 0.28, "spread": 0.35, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.35, "judge_b": 0.12, "mean": 0.22, "spread": 0.23, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209455" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:08.894Z" } }, { "rank": 199, "id": "st_20261024_606", "title": "Uncommon we:b 425 web libraries and components with a strong point of view", "url": "https://webdev.teknium.io/", "author": null, "site": "webdev.teknium.io", "published_at": "2026-10-23T13:27:46.000Z", "reading_minutes": 85, "depth": "three_judges", "format": "other", "summary": "Hrih News covers Uncommon we:b, a curated directory of 425 web libraries and components with a distinct visual point of view. The featured dither, pixel and ASCII section includes dithercn, a shadcn-registry component set that renders UI surfaces as ordered-dither cells, and Paper Shaders, zero-dependency WebGL effects configurable through props.…", "also_covered_by": [], "labels": { "content_type": "other", "project_stage": "not_a_project", "categories": [ "programming", "design" ] }, "score": { "total": 0.5951, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.495, "judge_a": 0.3, "judge_b": 0.42, "mean": 0.405, "spread": 0.195, "weight": 0.25 }, "reader_interest": { "jev": 0.76, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.76, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.563, "judge_a": 0.55, "judge_b": 0.58, "mean": 0.5644, "spread": 0.03, "weight": 0.1 }, "practical_relevance": { "jev": 0.783, "judge_a": 0.6, "judge_b": 0.68, "mean": 0.6875, "spread": 0.183, "weight": 0.15 }, "substantiated": { "jev": 0.937, "judge_a": 0.45, "judge_b": 0.86, "mean": 0.7489, "spread": 0.487, "weight": 0.1 }, "technical_depth": { "jev": 0.958, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8258, "spread": 0.258, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": 0.96, "mean": 0.8933, "spread": 0.16, "weight": 0.1 }, "promotional": { "jev": 0.1, "judge_a": 0.35, "judge_b": 0.18, "mean": 0.21, "spread": 0.25, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.25, "judge_b": 0.08, "mean": 0.22, "spread": 0.25, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50208661" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:09.007Z" } }, { "rank": 200, "id": "st_20261024_1204", "title": "List-Unsubscribe: Give recipients an alternative to marking email as spam", "url": "https://mxvalidation.com/blog/list-unsubscribe-header-trust-feedback", "author": null, "site": "MXValidation", "published_at": "2026-10-23T19:25:15.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "analysis_opinion", "summary": "An article on the List-Unsubscribe email header argues that making it easy for recipients to leave a mailing list reduces spam complaints and supports deliverability, since readers who can't find an exit often just hit \"report spam\" instead.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "growth" ] }, "score": { "total": 0.5916, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.475, "judge_a": 0.35, "judge_b": 0.48, "mean": 0.435, "spread": 0.13, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.55, "judge_b": 0.76, "mean": 0.6367, "spread": 0.21, "weight": 0.2 }, "audience_breadth": { "jev": 0.473, "judge_a": 0.5, "judge_b": 0.42, "mean": 0.4644, "spread": 0.08, "weight": 0.1 }, "practical_relevance": { "jev": 0.938, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8058, "spread": 0.238, "weight": 0.15 }, "substantiated": { "jev": 0.593, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.6578, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.888, "judge_a": 0.6, "judge_b": 0.55, "mean": 0.6792, "spread": 0.337, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.5, "judge_b": 0.95, "mean": 0.79, "spread": 0.45, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.05, "judge_b": 0.01, "mean": 0.0433, "spread": 0.06, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.15, "judge_b": 0.05, "mean": 0.12, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50214114" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:09.012Z" } }, { "rank": 201, "id": "st_20261024_1147", "title": "Kalshi's ETH perp trades 34 times its open interest a day", "url": "https://optimistfi.com/perps/kalshi-volume", "author": null, "site": "OptimistFi", "published_at": "2026-10-23T18:32:05.000Z", "reading_minutes": 8, "depth": "three_judges", "format": "original_research", "summary": "Kalshi's ETH perpetual futures market has turned over a median 34 times its open interest per day since October 13, against 1.5 times on Binance over the same days.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "data", "fintech" ] }, "score": { "total": 0.5907, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.438, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5392, "spread": 0.243, "weight": 0.25 }, "reader_interest": { "jev": 0.47, "judge_a": 0.8, "judge_b": 0.88, "mean": 0.7167, "spread": 0.41, "weight": 0.2 }, "audience_breadth": { "jev": 0.243, "judge_a": 0.5, "judge_b": 0.43, "mean": 0.3911, "spread": 0.257, "weight": 0.1 }, "practical_relevance": { "jev": 0.368, "judge_a": 0.4, "judge_b": 0.52, "mean": 0.4292, "spread": 0.153, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.7, "judge_b": 0.9, "mean": 0.8544, "spread": 0.263, "weight": 0.1 }, "technical_depth": { "jev": 0.975, "judge_a": 0.7, "judge_b": 0.86, "mean": 0.845, "spread": 0.275, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.9267, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": 0.03, "mean": 0.0467, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.41, "judge_a": 0.4, "judge_b": 0.12, "mean": 0.31, "spread": 0.29, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213417" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, not on front page", "checked_at": "2026-10-24T08:54:09.020Z" } }, { "rank": 202, "id": "st_20261024_1007", "title": "Meta delays its website-crawling terms after a publisher backlash", "url": "https://thenextweb.com/news/meta-delays-crawling-terms-publishers", "author": "Cristian Dina", "site": "The Next Web", "published_at": "2026-10-23T16:12:31.000Z", "first_published_at": "2026-10-23T16:12:31.000Z", "reading_minutes": 2, "depth": "three_judges", "format": "news_report", "summary": "Meta has delayed the start of new business terms from 30 October to 20 November after publishers said the terms would let it crawl their websites to build AI and advertising products.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "ai", "policy" ] }, "score": { "total": 0.5893, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.695, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5883, "spread": 0.245, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.72, "spread": 0.1, "weight": 0.2 }, "audience_breadth": { "jev": 0.467, "judge_a": 0.55, "judge_b": 0.7, "mean": 0.5722, "spread": 0.233, "weight": 0.1 }, "practical_relevance": { "jev": 0.618, "judge_a": 0.5, "judge_b": 0.58, "mean": 0.5658, "spread": 0.118, "weight": 0.15 }, "substantiated": { "jev": 0.533, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6011, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.535, "judge_a": 0.2, "judge_b": 0.35, "mean": 0.3617, "spread": 0.335, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": 0.94, "mean": 0.8833, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": 0.02, "mean": 0.0333, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.12, "judge_a": 0.15, "judge_b": 0.2, "mean": 0.1567, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.270Z" } }, { "rank": 203, "id": "st_20261024_1139", "title": "Cascache – A Go cache where slow reads can't undo invalidation", "url": "https://github.com/unkn0wn-root/cascache/", "author": null, "site": "GitHub", "published_at": "2026-10-23T18:40:34.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "release", "summary": "Cascache is a Go read-through cache library, released as a Show HN project, that stops slow reads from restoring stale data after an invalidation. Each key carries a fence token that Invalidate replaces, so a load started before the invalidation cannot store its old result.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming" ] }, "score": { "total": 0.5851, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.483, "judge_a": 0.3, "judge_b": 0.43, "mean": 0.4042, "spread": 0.183, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.55, "judge_b": 0.76, "mean": 0.65, "spread": 0.21, "weight": 0.2 }, "audience_breadth": { "jev": 0.327, "judge_a": 0.35, "judge_b": 0.34, "mean": 0.3389, "spread": 0.023, "weight": 0.1 }, "practical_relevance": { "jev": 0.915, "judge_a": 0.6, "judge_b": 0.74, "mean": 0.7517, "spread": 0.315, "weight": 0.15 }, "substantiated": { "jev": 0.8, "judge_a": 0.5, "judge_b": 0.79, "mean": 0.6967, "spread": 0.3, "weight": 0.1 }, "technical_depth": { "jev": 0.973, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.8175, "spread": 0.273, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.9, "judge_b": 0.98, "mean": 0.94, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.15, "judge_b": 0.14, "mean": 0.12, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.14, "judge_a": 0.2, "judge_b": 0.06, "mean": 0.1333, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213536" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:09.131Z" } }, { "rank": 204, "id": "st_20261024_869", "title": "Anthropic want work patterns to change", "url": "https://www.theguardian.com/technology/2026/oct/23/anthropic-report-ai-workplace-shifts", "author": "Ramin Skibba", "site": "the Guardian", "published_at": "2026-10-23T16:22:52.000Z", "reading_minutes": 5, "depth": "three_judges", "format": "news_report", "summary": "Anthropic published a report saying AI agents now do much of its work: entry-level technical hiring fell from nearly 30% to 9% since March 2025, and employees spend about 22% of their time managing Claude agents.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "programming", "ai", "startups" ] }, "score": { "total": 0.5673, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.693, "judge_a": 0.55, "judge_b": 0.72, "mean": 0.6542, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.85, "judge_b": 0.88, "mean": 0.8167, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.853, "judge_a": 0.7, "judge_b": 0.78, "mean": 0.7778, "spread": 0.153, "weight": 0.1 }, "practical_relevance": { "jev": 0.555, "judge_a": 0.45, "judge_b": 0.62, "mean": 0.5417, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.55, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.5067, "spread": 0.27, "weight": 0.1 }, "technical_depth": { "jev": 0.463, "judge_a": 0.2, "judge_b": 0.38, "mean": 0.3475, "spread": 0.263, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": 0.97, "mean": 0.86, "spread": 0.27, "weight": 0.1 }, "promotional": { "jev": 0.31, "judge_a": 0.2, "judge_b": 0.08, "mean": 0.1967, "spread": 0.23, "weight": -0.15 }, "overclaim": { "jev": 0.38, "judge_a": 0.35, "judge_b": 0.48, "mean": 0.4033, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50211493" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:09.134Z" } }, { "rank": 205, "id": "st_20261024_1228", "title": "CSE 452: Google's Introduction to Distributed System Design", "url": "https://courses.cs.washington.edu/courses/cse452/23wi/papers/yegge-platform-rant.html", "author": null, "site": "courses.cs.washington.edu", "published_at": "2026-10-23T19:01:02.000Z", "reading_minutes": 20, "depth": "three_judges", "format": "analysis_opinion", "summary": "Steve Yegge's 2011 internal rant, accidentally published on Google+, contrasts Google's engineering culture favorably against Amazon's, then recounts Jeff Bezos's 2002-era mandate that all teams expose functionality only through externalizable service interfaces.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "devops", "product" ] }, "score": { "total": 0.5641, "contested": true, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.718, "judge_a": 0.5, "judge_b": 0.72, "mean": 0.6458, "spread": 0.22, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.85, "judge_b": 0.94, "mean": 0.7967, "spread": 0.34, "weight": 0.2 }, "audience_breadth": { "jev": 0.843, "judge_a": 0.67, "judge_b": 0.78, "mean": 0.7644, "spread": 0.173, "weight": 0.1 }, "practical_relevance": { "jev": 0.66, "judge_a": 0.6, "judge_b": 0.83, "mean": 0.6967, "spread": 0.23, "weight": 0.15 }, "substantiated": { "jev": 0.147, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.3722, "spread": 0.473, "weight": 0.1 }, "technical_depth": { "jev": 0.595, "judge_a": 0.45, "judge_b": 0.73, "mean": 0.5917, "spread": 0.28, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.3, "judge_b": 0.98, "mean": 0.74, "spread": 0.68, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.02, "judge_b": 0.01, "mean": 0.0367, "spread": 0.07, "weight": -0.15 }, "overclaim": { "jev": 0.72, "judge_a": 0.45, "judge_b": 0.88, "mean": 0.6833, "spread": 0.43, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213812" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:09.144Z" } }, { "rank": 206, "id": "st_20261024_1764", "title": "BeNice! Closing Looping's Inference-Memory Gap via KV-Sharing", "url": "https://arxiv.org/abs/2610.31633", "author": "Hayden Prairie, Jacob M. Ryan, Taylor Berg-Kirkpatrick, Daniel Y. Fu", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:49:44.000Z", "depth": "two_judges", "format": "original_research", "summary": "The paper studies how to make looped language models share KV-caches across recurrence depths, which otherwise grow linearly in memory at inference. It proposes stochastic depth during pretraining to induce contractive residual dynamics, plus BeNice!, a post-training method that trains the first recurrence to produce KV-compatible residuals.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5605, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.62, "judge_a": 0.45, "judge_b": null, "mean": 0.535, "spread": 0.17, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.6, "judge_b": null, "mean": 0.6, "spread": 0, "weight": 0.2 }, "audience_breadth": { "jev": 0.473, "judge_a": 0.35, "judge_b": null, "mean": 0.4117, "spread": 0.123, "weight": 0.1 }, "practical_relevance": { "jev": 0.693, "judge_a": 0.5, "judge_b": null, "mean": 0.5962, "spread": 0.193, "weight": 0.15 }, "substantiated": { "jev": 0.6, "judge_a": 0.6, "judge_b": null, "mean": 0.6, "spread": 0, "weight": 0.1 }, "technical_depth": { "jev": 0.758, "judge_a": 0.75, "judge_b": null, "mean": 0.7537, "spread": 0.007, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.85, "judge_b": null, "mean": 0.895, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.05, "judge_b": null, "mean": 0.06, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.25, "judge_b": null, "mean": 0.265, "spread": 0.03, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.456Z" } }, { "rank": 207, "id": "st_20261024_522", "title": "The day PHP sites went offline on purpose", "url": "https://ohdear.app/uptime-statistics", "author": null, "site": "Oh Dear", "published_at": "2026-10-23T12:34:10.000Z", "reading_minutes": 15, "depth": "two_judges", "format": "original_research", "summary": "In October 2026, two critical PHP security vulnerabilities were disclosed before patches were ready, prompting many site owners to take their sites offline.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "security", "devops", "it" ] }, "score": { "total": 0.5604, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.643, "judge_a": 0.55, "judge_b": null, "mean": 0.5962, "spread": 0.092, "weight": 0.25 }, "reader_interest": { "jev": 0.68, "judge_a": 0.6, "judge_b": null, "mean": 0.64, "spread": 0.08, "weight": 0.2 }, "audience_breadth": { "jev": 0.657, "judge_a": 0.6, "judge_b": null, "mean": 0.6283, "spread": 0.057, "weight": 0.1 }, "practical_relevance": { "jev": 0.838, "judge_a": 0.5, "judge_b": null, "mean": 0.6688, "spread": 0.338, "weight": 0.15 }, "substantiated": { "jev": 0.957, "judge_a": 0.6, "judge_b": null, "mean": 0.7783, "spread": 0.357, "weight": 0.1 }, "technical_depth": { "jev": 0.833, "judge_a": 0.4, "judge_b": null, "mean": 0.6163, "spread": 0.433, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": null, "mean": 0.865, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.45, "judge_b": null, "mean": 0.27, "spread": 0.36, "weight": -0.15 }, "overclaim": { "jev": 0.47, "judge_a": 0.4, "judge_b": null, "mean": 0.435, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207993" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, not on front page", "checked_at": "2026-10-24T08:54:09.260Z" } }, { "rank": 208, "id": "st_20261024_1167", "title": "Anyctr – container runtime CLI gateway, use any container runtime in any project", "url": "https://github.com/pedrospdc/anyctr", "author": null, "site": "GitHub", "published_at": "2026-10-23T18:08:11.000Z", "reading_minutes": 7, "depth": "three_judges", "format": "release", "summary": "anyctr is an open-source command-line gateway that lets engineers use Docker-style commands (run, ps, logs, exec, and more) across Docker, Podman, and Apple's container runtime on macOS, Linux, and Windows.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "new_side_project", "categories": [ "programming", "devops" ] }, "score": { "total": 0.5603, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.483, "judge_a": 0.3, "judge_b": 0.34, "mean": 0.3742, "spread": 0.183, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.6, "judge_b": 0.62, "mean": 0.62, "spread": 0.04, "weight": 0.2 }, "audience_breadth": { "jev": 0.527, "judge_a": 0.4, "judge_b": 0.32, "mean": 0.4156, "spread": 0.207, "weight": 0.1 }, "practical_relevance": { "jev": 0.848, "judge_a": 0.7, "judge_b": 0.68, "mean": 0.7425, "spread": 0.167, "weight": 0.15 }, "substantiated": { "jev": 0.78, "judge_a": 0.6, "judge_b": 0.56, "mean": 0.6467, "spread": 0.22, "weight": 0.1 }, "technical_depth": { "jev": 0.98, "judge_a": 0.7, "judge_b": 0.82, "mean": 0.8333, "spread": 0.28, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.9, "judge_b": 0.96, "mean": 0.9333, "spread": 0.06, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.2, "judge_b": 0.12, "mean": 0.13, "spread": 0.13, "weight": -0.15 }, "overclaim": { "jev": 0.3, "judge_a": 0.2, "judge_b": 0.14, "mean": 0.2133, "spread": 0.16, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213084" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:09.265Z" } }, { "rank": 209, "id": "st_20261024_1589", "title": "Auditing the Hand-Written Inputs of Learned Access Control", "url": "https://arxiv.org/abs/2610.31388", "author": "Arsalan Jalil Sarghaleh, Mehdi Hazrati", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:53:37.000Z", "reading_minutes": 75, "depth": "two_judges", "format": "original_research", "summary": "This paper audits two hand-built components of learned access-control systems: the hard-deny clauses that a shield enforces and the region map that routes requests between mined rules and a learner.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "security" ] }, "score": { "total": 0.5597, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.573, "judge_a": 0.35, "judge_b": null, "mean": 0.4613, "spread": 0.223, "weight": 0.25 }, "reader_interest": { "jev": 0.57, "judge_a": 0.4, "judge_b": null, "mean": 0.485, "spread": 0.17, "weight": 0.2 }, "audience_breadth": { "jev": 0.443, "judge_a": 0.3, "judge_b": null, "mean": 0.3717, "spread": 0.143, "weight": 0.1 }, "practical_relevance": { "jev": 0.715, "judge_a": 0.4, "judge_b": null, "mean": 0.5575, "spread": 0.315, "weight": 0.15 }, "substantiated": { "jev": 0.99, "judge_a": 0.85, "judge_b": null, "mean": 0.92, "spread": 0.14, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.8, "judge_b": null, "mean": 0.8988, "spread": 0.198, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.8, "judge_b": null, "mean": 0.845, "spread": 0.09, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": null, "mean": 0.045, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.2, "judge_b": null, "mean": 0.22, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.561Z" } }, { "rank": 210, "id": "st_20261024_1604", "title": "Would an LLM Match into a Residency? Uncertainty, Reasoning and Image Use of Open Language Models on the Spanish MIR Examination", "url": "https://arxiv.org/abs/2610.31586", "author": "Miguel Fernandez-de-Retana, Unai Zulaika, Ruben Sanchez-Corcuera et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:23:11.000Z", "reading_minutes": 32, "depth": "two_judges", "format": "original_research", "summary": "This paper evaluates four open language models (8 to 27 billion parameters) on 1,552 questions from eight editions of Spain's MIR residency examination, scored under the exam's own penalty and allocation rules.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5593, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.6, "judge_a": 0.35, "judge_b": null, "mean": 0.475, "spread": 0.25, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.4, "judge_b": null, "mean": 0.505, "spread": 0.21, "weight": 0.2 }, "audience_breadth": { "jev": 0.503, "judge_a": 0.3, "judge_b": null, "mean": 0.4017, "spread": 0.203, "weight": 0.1 }, "practical_relevance": { "jev": 0.705, "judge_a": 0.45, "judge_b": null, "mean": 0.5775, "spread": 0.255, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.8, "judge_b": null, "mean": 0.89, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.7, "judge_b": null, "mean": 0.845, "spread": 0.29, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": null, "mean": 0.865, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": null, "mean": 0.045, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.3, "judge_b": null, "mean": 0.27, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.602Z" } }, { "rank": 211, "id": "st_20261024_1174", "title": "I want my Wednesday back", "url": "https://blog.luitjes.it/posts/i-want-my-wednesday-back/", "author": null, "site": "blog.luitjes.it", "published_at": "2026-10-23T18:02:41.000Z", "reading_minutes": 3, "depth": "two_judges", "format": "analysis_opinion", "summary": "A blog post argues that php2shell is a serious PHP vulnerability but unlikely to be mass-exploited quickly, since the public proof-of-concept needs four consecutive requests to the same PHP worker, exact build knowledge, and tens of thousands of brute-force attempts that would likely trigger detection and blocking.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "security" ] }, "score": { "total": 0.5589, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.523, "judge_a": 0.45, "judge_b": null, "mean": 0.4863, "spread": 0.072, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.7, "judge_b": null, "mean": 0.655, "spread": 0.09, "weight": 0.2 }, "audience_breadth": { "jev": 0.53, "judge_a": 0.6, "judge_b": null, "mean": 0.565, "spread": 0.07, "weight": 0.1 }, "practical_relevance": { "jev": 0.593, "judge_a": 0.65, "judge_b": null, "mean": 0.6213, "spread": 0.057, "weight": 0.15 }, "substantiated": { "jev": 0.533, "judge_a": 0.6, "judge_b": null, "mean": 0.5667, "spread": 0.067, "weight": 0.1 }, "technical_depth": { "jev": 0.78, "judge_a": 0.7, "judge_b": null, "mean": 0.74, "spread": 0.08, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.5, "judge_b": null, "mean": 0.695, "spread": 0.39, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.05, "judge_b": null, "mean": 0.07, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.2, "judge_b": null, "mean": 0.22, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 4, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213001" }, "lobsters": null, "line": "Elsewhere: HN 4 pts, not on front page", "checked_at": "2026-10-24T08:54:09.394Z" } }, { "rank": 212, "id": "st_20261024_1379", "title": "envoyproxy/envoy v1.39.5", "url": "https://github.com/envoyproxy/envoy/releases/tag/v1.39.5", "author": "publish-envoy[bot]", "site": "GitHub", "published_at": "2026-10-23T01:01:48.000Z", "first_published_at": "2026-10-23T01:01:48.000Z", "reading_minutes": 1, "depth": "two_judges", "format": "release", "summary": "Envoy v1.39.5 ships fixes for an upstream compromise of several eMudhra intermediate CA private keys, removing trust in those CAs from the Docker images, the webpki-roots and rustls-webpki crates used by Hickory DNS, and two roots (AAA Certificate Services, emSign ECC Root CA - G3) from Envoy Mobile's built-in bundle.…", "also_covered_by": [ { "site": "GitHub", "url": "https://github.com/envoyproxy/envoy/releases/tag/v1.40.1", "title": "envoyproxy/envoy v1.40.1" } ], "labels": { "content_type": "release", "project_stage": "product_release", "categories": [ "programming", "security", "devops" ] }, "score": { "total": 0.5585, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.665, "judge_a": 0.35, "judge_b": null, "mean": 0.5075, "spread": 0.315, "weight": 0.25 }, "reader_interest": { "jev": 0.59, "judge_a": 0.4, "judge_b": null, "mean": 0.495, "spread": 0.19, "weight": 0.2 }, "audience_breadth": { "jev": 0.543, "judge_a": 0.4, "judge_b": null, "mean": 0.4717, "spread": 0.143, "weight": 0.1 }, "practical_relevance": { "jev": 0.868, "judge_a": 0.6, "judge_b": null, "mean": 0.7338, "spread": 0.268, "weight": 0.15 }, "substantiated": { "jev": 0.753, "judge_a": 0.5, "judge_b": null, "mean": 0.6267, "spread": 0.253, "weight": 0.1 }, "technical_depth": { "jev": 0.68, "judge_a": 0.35, "judge_b": null, "mean": 0.515, "spread": 0.33, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.7, "judge_b": null, "mean": 0.815, "spread": 0.23, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.03, "judge_b": null, "mean": 0.04, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.09, "judge_a": 0.1, "judge_b": null, "mean": 0.095, "spread": 0.01, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.712Z" } }, { "rank": 213, "id": "st_20261024_78", "title": "The Token Exchange", "url": "https://fusionauth.io/blog/token-exchange-thoughts", "author": null, "site": "FusionAuth", "published_at": "2026-10-23T02:46:23.000Z", "reading_minutes": 8, "depth": "two_judges", "format": "analysis_opinion", "summary": "Hrih News summary: This article explains OAuth 2.0 token exchange (RFC 8693), a standard that lets a service trade a token it holds for a narrower token scoped to a specific downstream service, preserving user identity while limiting blast radius.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming", "ai", "security" ] }, "score": { "total": 0.5584, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.59, "judge_a": 0.3, "judge_b": null, "mean": 0.445, "spread": 0.29, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.6, "judge_b": null, "mean": 0.66, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.62, "judge_a": 0.5, "judge_b": null, "mean": 0.56, "spread": 0.12, "weight": 0.1 }, "practical_relevance": { "jev": 0.813, "judge_a": 0.7, "judge_b": null, "mean": 0.7563, "spread": 0.113, "weight": 0.15 }, "substantiated": { "jev": 0.45, "judge_a": 0.45, "judge_b": null, "mean": 0.45, "spread": 0, "weight": 0.1 }, "technical_depth": { "jev": 0.855, "judge_a": 0.7, "judge_b": null, "mean": 0.7775, "spread": 0.155, "weight": 0.1 }, "new_information": { "jev": 0.87, "judge_a": 0.4, "judge_b": null, "mean": 0.635, "spread": 0.47, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.15, "judge_b": null, "mean": 0.1, "spread": 0.1, "weight": -0.15 }, "overclaim": { "jev": 0.14, "judge_a": 0.2, "judge_b": null, "mean": 0.17, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203909" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:09.581Z" } }, { "rank": 214, "id": "st_20261024_1066", "title": "Containers_basics – Basic Barebones Zero-Dependency Linux Containers in C", "url": "https://github.com/sharadg/containers_basics", "author": null, "site": "GitHub", "published_at": "2026-10-23T17:55:57.000Z", "reading_minutes": 13, "depth": "two_judges", "format": "new_projects", "summary": "The Containers_basics project is a zero-dependency C tutorial that builds Linux containers from the ground up, using fork/exec, clone-based namespaces (PID, UTS, mount, and others), cgroups for memory limits, and unshare/chroot with an exported Docker rootfs.", "also_covered_by": [], "labels": { "content_type": "tutorial", "project_stage": "new_side_project", "categories": [ "programming", "devops" ] }, "score": { "total": 0.5584, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.46, "judge_a": 0.3, "judge_b": null, "mean": 0.38, "spread": 0.16, "weight": 0.25 }, "reader_interest": { "jev": 0.35, "judge_a": 0.7, "judge_b": null, "mean": 0.525, "spread": 0.35, "weight": 0.2 }, "audience_breadth": { "jev": 0.453, "judge_a": 0.5, "judge_b": null, "mean": 0.4767, "spread": 0.047, "weight": 0.1 }, "practical_relevance": { "jev": 0.705, "judge_a": 0.6, "judge_b": null, "mean": 0.6525, "spread": 0.105, "weight": 0.15 }, "substantiated": { "jev": 0.95, "judge_a": 0.7, "judge_b": null, "mean": 0.825, "spread": 0.25, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.9, "judge_b": null, "mean": 0.9488, "spread": 0.098, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.5, "judge_b": null, "mean": 0.715, "spread": 0.43, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.05, "judge_b": null, "mean": 0.04, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.15, "judge_b": null, "mean": 0.2, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212899" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:09.686Z" } }, { "rank": 215, "id": "st_20261024_1689", "title": "A Quadratic Lens on Muon's Polar Update: Benefits, Pitfalls, and the Role of Newton-Schulz", "url": "https://arxiv.org/abs/2610.30280", "author": "Egor Shulgin, Sam Laing, Antonio Orvieto, Peter Richtárik", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T18:01:04.000Z", "depth": "two_judges", "format": "original_research", "summary": "This paper analyzes PolarGD, the orthogonalized update at the core of the Muon optimizer, using closed-form matrix quadratics where optimizers can be compared without learning-rate tuning.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5578, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.58, "judge_a": 0.4, "judge_b": null, "mean": 0.49, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.52, "judge_a": 0.6, "judge_b": null, "mean": 0.56, "spread": 0.08, "weight": 0.2 }, "audience_breadth": { "jev": 0.433, "judge_a": 0.4, "judge_b": null, "mean": 0.4167, "spread": 0.033, "weight": 0.1 }, "practical_relevance": { "jev": 0.575, "judge_a": 0.55, "judge_b": null, "mean": 0.5625, "spread": 0.025, "weight": 0.15 }, "substantiated": { "jev": 0.663, "judge_a": 0.6, "judge_b": null, "mean": 0.6317, "spread": 0.063, "weight": 0.1 }, "technical_depth": { "jev": 0.853, "judge_a": 0.85, "judge_b": null, "mean": 0.8513, "spread": 0.003, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.8, "judge_b": null, "mean": 0.865, "spread": 0.13, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.02, "judge_b": null, "mean": 0.035, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.2, "judge_b": null, "mean": 0.215, "spread": 0.03, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.923Z" } }, { "rank": 216, "id": "st_20261024_1465", "title": "The Value of Memory in Language Agents", "url": "https://arxiv.org/abs/2610.31303", "author": "Zehao Lin, Erenkai Zhu, Helan Hu et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:01:31.000Z", "reading_minutes": 43, "depth": "two_judges", "format": "original_research", "summary": "This paper proposes a decision-theoretic framework for valuing memory in language agents, tracing value from raw history through stored state, usable recovery and costly revision. It uses avoided task loss as the common unit, supported by simulation counterexamples and diagnostics on LoCoMo and longitudinal cases.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5576, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.643, "judge_a": 0.35, "judge_b": null, "mean": 0.4963, "spread": 0.293, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.55, "judge_b": null, "mean": 0.545, "spread": 0.01, "weight": 0.2 }, "audience_breadth": { "jev": 0.573, "judge_a": 0.5, "judge_b": null, "mean": 0.5367, "spread": 0.073, "weight": 0.1 }, "practical_relevance": { "jev": 0.698, "judge_a": 0.5, "judge_b": null, "mean": 0.5988, "spread": 0.198, "weight": 0.15 }, "substantiated": { "jev": 0.907, "judge_a": 0.6, "judge_b": null, "mean": 0.7533, "spread": 0.307, "weight": 0.1 }, "technical_depth": { "jev": 0.975, "judge_a": 0.7, "judge_b": null, "mean": 0.8375, "spread": 0.275, "weight": 0.1 }, "new_information": { "jev": 0.86, "judge_a": 0.6, "judge_b": null, "mean": 0.73, "spread": 0.26, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": null, "mean": 0.05, "spread": 0, "weight": -0.15 }, "overclaim": { "jev": 0.23, "judge_a": 0.35, "judge_b": null, "mean": 0.29, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:09.887Z" } }, { "rank": 217, "id": "st_20261024_1152", "title": "Greece enacts social media ban for kids under 15", "url": "https://www.engadget.com/2295143/greece-enacts-social-media-ban-for-kids-under-15/", "author": "Steve Dent", "site": "Engadget", "published_at": "2026-10-23T18:24:26.000Z", "reading_minutes": 1, "depth": "two_judges", "format": "news_report", "summary": "Greece's parliament passed a law banning social media accounts for children under 15, effective January 1, 2027. It covers platforms like TikTok, Instagram, Facebook and X, requires proportionate age verification, and enforces compliance through fines under the EU's Digital Services Act.", "also_covered_by": [ { "site": "Silicon Republic", "url": "https://www.siliconrepublic.com/business/greek-parliament-greenlights-under-15-social-media-ban", "title": "Greek parliament greenlights under-15 social media ban" }, { "site": "Slashdot", "url": "https://tech.slashdot.org/story/26/10/23/1928227/greece-enacts-social-media-ban-for-kids-under-15", "title": "Greece Enacts Social Media Ban For Kids Under 15" }, { "site": "reuters.com", "url": "https://www.reuters.com/legal/litigation/greece-passes-social-media-ban-children-under-15-2026-10-23/", "title": "Greece passes social media ban for children under 15" } ], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "policy" ] }, "score": { "total": 0.5574, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.68, "judge_a": 0.55, "judge_b": null, "mean": 0.615, "spread": 0.13, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.7, "judge_b": null, "mean": 0.62, "spread": 0.16, "weight": 0.2 }, "audience_breadth": { "jev": 0.593, "judge_a": 0.6, "judge_b": null, "mean": 0.5967, "spread": 0.007, "weight": 0.1 }, "practical_relevance": { "jev": 0.685, "judge_a": 0.45, "judge_b": null, "mean": 0.5675, "spread": 0.235, "weight": 0.15 }, "substantiated": { "jev": 0.44, "judge_a": 0.45, "judge_b": null, "mean": 0.445, "spread": 0.01, "weight": 0.1 }, "technical_depth": { "jev": 0.578, "judge_a": 0.3, "judge_b": null, "mean": 0.4388, "spread": 0.278, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": null, "mean": 0.855, "spread": 0.11, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": null, "mean": 0.05, "spread": 0, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.2, "judge_b": null, "mean": 0.21, "spread": 0.02, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 6, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213328" }, "lobsters": null, "line": "Elsewhere: HN 6 pts, not on front page", "checked_at": "2026-10-24T08:54:09.816Z" } }, { "rank": 218, "id": "st_20261024_1595", "title": "How Depth Growth and Looping Shape In-Context Learning in Language Models", "url": "https://arxiv.org/abs/2610.31421", "author": "Ferdinand Kapl, Kaitlin Maile, Stefan Bauer, Johannes von Oswald", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:21:22.000Z", "reading_minutes": 62, "depth": "two_judges", "format": "original_research", "summary": "The paper examines whether reusing computation across depth, via depth growing or looping in language models, favors in-context learning. Experiments on pretrained models up to 3B parameters and controlled tasks suggest grown and looped models rely more on context, acquire and retain ICL more readily than a dense baseline, and…", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5573, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.603, "judge_a": 0.45, "judge_b": null, "mean": 0.5263, "spread": 0.153, "weight": 0.25 }, "reader_interest": { "jev": 0.5, "judge_a": 0.55, "judge_b": null, "mean": 0.525, "spread": 0.05, "weight": 0.2 }, "audience_breadth": { "jev": 0.537, "judge_a": 0.4, "judge_b": null, "mean": 0.4683, "spread": 0.137, "weight": 0.1 }, "practical_relevance": { "jev": 0.52, "judge_a": 0.35, "judge_b": null, "mean": 0.435, "spread": 0.17, "weight": 0.15 }, "substantiated": { "jev": 0.95, "judge_a": 0.8, "judge_b": null, "mean": 0.875, "spread": 0.15, "weight": 0.1 }, "technical_depth": { "jev": 0.948, "judge_a": 0.85, "judge_b": null, "mean": 0.8988, "spread": 0.098, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.7, "judge_b": null, "mean": 0.8, "spread": 0.2, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": null, "mean": 0.035, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.3, "judge_b": null, "mean": 0.29, "spread": 0.02, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.031Z" } }, { "rank": 219, "id": "st_20261024_1602", "title": "When to Follow the Teacher: Advantage-Based On-Policy Distillation", "url": "https://arxiv.org/abs/2610.31563", "author": "Mahdi Sabbaghi, George Pappas, Hamed Hassani, Adel Javanmard", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:05:21.000Z", "reading_minutes": 37, "depth": "two_judges", "format": "original_research", "summary": "AOPD is a proposed improvement to on-policy distillation, where a student model learns from a teacher along its own generated trajectories. Standard methods make the student imitate the teacher at every step, even when that guidance hurts success.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5572, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.603, "judge_a": 0.4, "judge_b": null, "mean": 0.5013, "spread": 0.203, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.6, "judge_b": null, "mean": 0.57, "spread": 0.06, "weight": 0.2 }, "audience_breadth": { "jev": 0.537, "judge_a": 0.5, "judge_b": null, "mean": 0.5183, "spread": 0.037, "weight": 0.1 }, "practical_relevance": { "jev": 0.623, "judge_a": 0.55, "judge_b": null, "mean": 0.5863, "spread": 0.073, "weight": 0.15 }, "substantiated": { "jev": 0.947, "judge_a": 0.6, "judge_b": null, "mean": 0.7733, "spread": 0.347, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.8, "judge_b": null, "mean": 0.895, "spread": 0.19, "weight": 0.1 }, "new_information": { "jev": 0.88, "judge_a": 0.8, "judge_b": null, "mean": 0.84, "spread": 0.08, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": null, "mean": 0.05, "spread": 0, "weight": -0.15 }, "overclaim": { "jev": 0.52, "judge_a": 0.35, "judge_b": null, "mean": 0.435, "spread": 0.17, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.119Z" } }, { "rank": 220, "id": "st_20261024_1679", "title": "SQuAT: Spherical Lattice Vector Quantization-Aware Training", "url": "https://arxiv.org/abs/2610.31412", "author": "Sebastian Loeschcke, Ivan Koryakovskiy, Tycho F. A. van der Ouderaa et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T15:14:44.000Z", "depth": "two_judges", "format": "original_research", "summary": "SQuAT is a quantization-aware training method for low-bit large language models that splits each weight into a direction and a magnitude, quantizing directions with an E8-derived spherical codebook and gains with scalar quantizers.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5567, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.648, "judge_a": 0.5, "judge_b": null, "mean": 0.5738, "spread": 0.147, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.6, "judge_b": null, "mean": 0.61, "spread": 0.02, "weight": 0.2 }, "audience_breadth": { "jev": 0.44, "judge_a": 0.4, "judge_b": null, "mean": 0.42, "spread": 0.04, "weight": 0.1 }, "practical_relevance": { "jev": 0.588, "judge_a": 0.55, "judge_b": null, "mean": 0.5688, "spread": 0.037, "weight": 0.15 }, "substantiated": { "jev": 0.543, "judge_a": 0.6, "judge_b": null, "mean": 0.5717, "spread": 0.057, "weight": 0.1 }, "technical_depth": { "jev": 0.78, "judge_a": 0.8, "judge_b": null, "mean": 0.79, "spread": 0.02, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": null, "mean": 0.87, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": null, "mean": 0.055, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.33, "judge_a": 0.35, "judge_b": null, "mean": 0.34, "spread": 0.02, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.186Z" } }, { "rank": 221, "id": "st_20261024_1507", "title": "TerraTransfer++: Demonstration-Free End-to-End Planner", "url": "https://arxiv.org/abs/2610.30319", "author": "Zikang Xiong, Weixin Li, Zhouchonghao Wu et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T18:31:28.000Z", "reading_minutes": 72, "depth": "two_judges", "format": "original_research", "summary": "TerraTransfer++ proposes training end-to-end autonomous driving policies without human demonstrations. A driving policy is first trained by multi-agent self-play in a vectorized simulator, then distilled into a trajectory teacher, and finally aligned to a frozen pretrained vision backbone through lightweight adapters.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5564, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.715, "judge_a": 0.45, "judge_b": null, "mean": 0.5825, "spread": 0.265, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.55, "judge_b": null, "mean": 0.625, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.417, "judge_a": 0.4, "judge_b": null, "mean": 0.4083, "spread": 0.017, "weight": 0.1 }, "practical_relevance": { "jev": 0.628, "judge_a": 0.45, "judge_b": null, "mean": 0.5387, "spread": 0.177, "weight": 0.15 }, "substantiated": { "jev": 0.843, "judge_a": 0.6, "judge_b": null, "mean": 0.7217, "spread": 0.243, "weight": 0.1 }, "technical_depth": { "jev": 0.99, "judge_a": 0.9, "judge_b": null, "mean": 0.945, "spread": 0.09, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": null, "mean": 0.86, "spread": 0.12, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.1, "judge_b": null, "mean": 0.085, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.56, "judge_a": 0.45, "judge_b": null, "mean": 0.505, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.255Z" } }, { "rank": 222, "id": "st_20261024_1537", "title": "ULME: Predictive Coupling for Mutually Reinforcing Embedding and Language Modeling", "url": "https://arxiv.org/abs/2610.30705", "author": "Yang Liu, Zhiqiang Yuan, Shu-Dong Huang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T04:23:07.000Z", "reading_minutes": 37, "depth": "two_judges", "format": "original_research", "summary": "ULME is a post-training method that trains a single causal decoder LLM for both text embedding and answer generation by inserting learnable slot tokens between query and answer and coupling them with three losses: contrastive, language modeling, and a stop-gradient predictive alignment.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5562, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.68, "judge_a": 0.35, "judge_b": null, "mean": 0.515, "spread": 0.33, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.45, "judge_b": null, "mean": 0.555, "spread": 0.21, "weight": 0.2 }, "audience_breadth": { "jev": 0.597, "judge_a": 0.45, "judge_b": null, "mean": 0.5233, "spread": 0.147, "weight": 0.1 }, "practical_relevance": { "jev": 0.718, "judge_a": 0.5, "judge_b": null, "mean": 0.6088, "spread": 0.218, "weight": 0.15 }, "substantiated": { "jev": 0.927, "judge_a": 0.6, "judge_b": null, "mean": 0.7633, "spread": 0.327, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.75, "judge_b": null, "mean": 0.86, "spread": 0.22, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": null, "mean": 0.795, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.1, "judge_b": null, "mean": 0.08, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.41, "judge_a": 0.35, "judge_b": null, "mean": 0.38, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.300Z" } }, { "rank": 223, "id": "st_20261024_1542", "title": "VLTest: Finding Joint Image-Text Failures in Vision-Language Models", "url": "https://arxiv.org/abs/2610.30794", "author": "Bin Duan, Jintao Lin, Guowei Yang", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T06:19:39.000Z", "reading_minutes": 64, "depth": "two_judges", "format": "original_research", "summary": "VLTest is a black-box testing approach for vision-language models that addresses two problems: an output change is not necessarily a failure, since mutations can alter the correct answer, and joint failures require matched single-modality controls to be attributed to the combination.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5562, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.595, "judge_a": 0.4, "judge_b": null, "mean": 0.4975, "spread": 0.195, "weight": 0.25 }, "reader_interest": { "jev": 0.49, "judge_a": 0.45, "judge_b": null, "mean": 0.47, "spread": 0.04, "weight": 0.2 }, "audience_breadth": { "jev": 0.55, "judge_a": 0.35, "judge_b": null, "mean": 0.45, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.665, "judge_a": 0.45, "judge_b": null, "mean": 0.5575, "spread": 0.215, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.8, "judge_b": null, "mean": 0.8933, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.8, "judge_b": null, "mean": 0.8963, "spread": 0.193, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.6, "judge_b": null, "mean": 0.76, "spread": 0.32, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": null, "mean": 0.045, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.22, "judge_a": 0.3, "judge_b": null, "mean": 0.26, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.392Z" } }, { "rank": 224, "id": "st_20261024_1791", "title": "Memory before Goal: Role of Pre-History Memory for Humanoid Loco-Manipulation", "url": "https://arxiv.org/abs/2610.31166", "author": "Chenyuan Chen, Ran Ding, Lechen Zhang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T12:24:25.000Z", "depth": "two_judges", "format": "original_research", "summary": "WholeBodyMem addresses a gap in humanoid robot control: vision-language-action policies see only the current frame, so cues vanish when objects are hidden or leave view during walking and turning.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5557, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.64, "judge_a": 0.45, "judge_b": null, "mean": 0.545, "spread": 0.19, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.55, "judge_b": null, "mean": 0.585, "spread": 0.07, "weight": 0.2 }, "audience_breadth": { "jev": 0.407, "judge_a": 0.35, "judge_b": null, "mean": 0.3783, "spread": 0.057, "weight": 0.1 }, "practical_relevance": { "jev": 0.55, "judge_a": 0.5, "judge_b": null, "mean": 0.525, "spread": 0.05, "weight": 0.15 }, "substantiated": { "jev": 0.787, "judge_a": 0.6, "judge_b": null, "mean": 0.6933, "spread": 0.187, "weight": 0.1 }, "technical_depth": { "jev": 0.745, "judge_a": 0.7, "judge_b": null, "mean": 0.7225, "spread": 0.045, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": null, "mean": 0.915, "spread": 0.03, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.1, "judge_b": null, "mean": 0.085, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.3, "judge_b": null, "mean": 0.23, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.481Z" } }, { "rank": 225, "id": "st_20261024_725", "title": "Profiling Org->HTML formal grammar and optimizing it: orgwebalchemy", "url": "https://jointhefreeworld.org/blog/articles/lisps/orgwebalchemy-benchmarking-profiling-and-org-support/", "author": "Josep Bigorra - jointhefreeworld", "site": "jointhefreeworld", "published_at": "2026-10-23T14:01:05.000Z", "reading_minutes": 25, "depth": "two_judges", "format": "engineering_writeup", "summary": "OrgWebAlchemy, a Guile library that parses Org mode with a Parsing Expression Grammar and renders it to HTML, Markdown, or SXML, has been optimized and extended.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "early_project", "categories": [ "programming" ] }, "score": { "total": 0.5553, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.48, "judge_a": 0.3, "judge_b": null, "mean": 0.39, "spread": 0.18, "weight": 0.25 }, "reader_interest": { "jev": 0.65, "judge_a": 0.55, "judge_b": null, "mean": 0.6, "spread": 0.1, "weight": 0.2 }, "audience_breadth": { "jev": 0.367, "judge_a": 0.25, "judge_b": null, "mean": 0.3083, "spread": 0.117, "weight": 0.1 }, "practical_relevance": { "jev": 0.778, "judge_a": 0.35, "judge_b": null, "mean": 0.5638, "spread": 0.428, "weight": 0.15 }, "substantiated": { "jev": 0.993, "judge_a": 0.8, "judge_b": null, "mean": 0.8967, "spread": 0.193, "weight": 0.1 }, "technical_depth": { "jev": 1, "judge_a": 0.8, "judge_b": null, "mean": 0.9, "spread": 0.2, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": null, "mean": 0.855, "spread": 0.11, "weight": 0.1 }, "promotional": { "jev": 0.03, "judge_a": 0.15, "judge_b": null, "mean": 0.09, "spread": 0.12, "weight": -0.15 }, "overclaim": { "jev": 0.14, "judge_a": 0.25, "judge_b": null, "mean": 0.195, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50209148" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:10.388Z" } }, { "rank": 226, "id": "st_20261024_1092", "title": "Chroma Cloud wind-down timeline", "url": "https://www.trychroma.com/cloud-wind-down", "author": null, "site": "trychroma.com", "published_at": "2026-10-23T17:22:17.000Z", "reading_minutes": 1, "depth": "two_judges", "format": "news_report", "summary": "Chroma Cloud, the hosted version of the ChromaDB vector database, is shutting down over 60 days from October 23, 2026. Writes are disabled on November 22 and reads on December 22, so customers must migrate data and workloads, while open-source ChromaDB continues under Apache 2.0.", "also_covered_by": [ { "site": "trychroma.com", "url": "https://www.trychroma.com/blog/chroma-joins-replit", "title": "Chroma Is Joining Replit" } ], "labels": { "content_type": "news_report", "project_stage": "established_project", "categories": [ "ai", "devops", "data" ] }, "score": { "total": 0.5553, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.507, "judge_a": 0.4, "judge_b": null, "mean": 0.4538, "spread": 0.107, "weight": 0.25 }, "reader_interest": { "jev": 0.61, "judge_a": 0.7, "judge_b": null, "mean": 0.655, "spread": 0.09, "weight": 0.2 }, "audience_breadth": { "jev": 0.11, "judge_a": 0.4, "judge_b": null, "mean": 0.255, "spread": 0.29, "weight": 0.1 }, "practical_relevance": { "jev": 0.903, "judge_a": 0.8, "judge_b": null, "mean": 0.8513, "spread": 0.102, "weight": 0.15 }, "substantiated": { "jev": 0.483, "judge_a": 0.6, "judge_b": null, "mean": 0.5417, "spread": 0.117, "weight": 0.1 }, "technical_depth": { "jev": 0.53, "judge_a": 0.3, "judge_b": null, "mean": 0.415, "spread": 0.23, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": null, "mean": 0.86, "spread": 0.12, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.1, "judge_b": null, "mean": 0.08, "spread": 0.04, "weight": -0.15 }, "overclaim": { "jev": 0.06, "judge_a": 0.1, "judge_b": null, "mean": 0.08, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50212417" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:10.424Z" } }, { "rank": 227, "id": "st_20261024_953", "title": "Researchers weave multifunctional touchscreen into sweater sleeve", "url": "https://www.heise.de/en/news/Researchers-weave-multifunctional-touchscreen-into-sweater-sleeve-11498217.html", "author": "heise online", "site": "heise online English", "published_at": "2026-10-23T12:19:00.000Z", "first_published_at": "2026-10-23T12:19:00.000Z", "reading_minutes": 3, "depth": "two_judges", "format": "news_report", "summary": "NUS researchers wove a flexible, washable, breathable touchscreen into a sweater sleeve, built from 54 touch-sensitive LED fibers with a 96 × 54 pixel resolution. The fibers, made by thermal drawing with laser-programmed solder connections, also carry microphones, sensors, and speakers.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "not_a_project", "categories": [ "hardware" ] }, "score": { "total": 0.5552, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.595, "judge_a": 0.45, "judge_b": null, "mean": 0.5225, "spread": 0.145, "weight": 0.25 }, "reader_interest": { "jev": 0.73, "judge_a": 0.6, "judge_b": null, "mean": 0.665, "spread": 0.13, "weight": 0.2 }, "audience_breadth": { "jev": 0.47, "judge_a": 0.4, "judge_b": null, "mean": 0.435, "spread": 0.07, "weight": 0.1 }, "practical_relevance": { "jev": 0.53, "judge_a": 0.35, "judge_b": null, "mean": 0.44, "spread": 0.18, "weight": 0.15 }, "substantiated": { "jev": 0.727, "judge_a": 0.7, "judge_b": null, "mean": 0.7133, "spread": 0.027, "weight": 0.1 }, "technical_depth": { "jev": 0.765, "judge_a": 0.6, "judge_b": null, "mean": 0.6825, "spread": 0.165, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.9, "judge_b": null, "mean": 0.92, "spread": 0.04, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.1, "judge_b": null, "mean": 0.095, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.17, "judge_a": 0.3, "judge_b": null, "mean": 0.235, "spread": 0.13, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.698Z" } }, { "rank": 228, "id": "st_20261024_1547", "title": "Separate Field Scoring Protects Unused Tools When Agents Pool Tool-Use Experience", "url": "https://arxiv.org/abs/2610.30867", "author": "Abhijit Chakraborty", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T07:45:24.000Z", "reading_minutes": 23, "depth": "two_judges", "format": "original_research", "summary": "Researchers studied how pooled usage records affect tool retrieval for AI agents sharing a tool catalog, measuring separately tools with no usage (held-out) and tools with usage.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5552, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.565, "judge_a": 0.35, "judge_b": null, "mean": 0.4575, "spread": 0.215, "weight": 0.25 }, "reader_interest": { "jev": 0.56, "judge_a": 0.45, "judge_b": null, "mean": 0.505, "spread": 0.11, "weight": 0.2 }, "audience_breadth": { "jev": 0.487, "judge_a": 0.4, "judge_b": null, "mean": 0.4433, "spread": 0.087, "weight": 0.1 }, "practical_relevance": { "jev": 0.718, "judge_a": 0.5, "judge_b": null, "mean": 0.6088, "spread": 0.218, "weight": 0.15 }, "substantiated": { "jev": 0.987, "judge_a": 0.7, "judge_b": null, "mean": 0.8433, "spread": 0.287, "weight": 0.1 }, "technical_depth": { "jev": 0.998, "judge_a": 0.8, "judge_b": null, "mean": 0.8988, "spread": 0.198, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.7, "judge_b": null, "mean": 0.795, "spread": 0.19, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": null, "mean": 0.045, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.32, "judge_a": 0.25, "judge_b": null, "mean": 0.285, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.683Z" } }, { "rank": 229, "id": "st_20261024_1255", "title": "Kimi K3.1 Pro puts open-source AI a month behind the best", "url": "https://www.modelcompare.ai/blog/china-open-source-ai", "author": "ModelCompare · 23 October 2026", "site": "ModelCompare", "published_at": "2026-10-23T20:15:27.000Z", "reading_minutes": 5, "depth": "two_judges", "format": "news_report", "summary": "Kimi released K3.1 Pro on 23 October 2026, scoring 55.1 on the Artificial Analysis Intelligence Index, 2.5 points behind Claude Opus 5.5 and 8.8 points ahead of the previous best open model. Kimi says the weights will be downloadable by 4 November.", "also_covered_by": [], "labels": { "content_type": "news_report", "project_stage": "product_release", "categories": [ "ai" ] }, "score": { "total": 0.5551, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.675, "judge_a": 0.55, "judge_b": null, "mean": 0.6125, "spread": 0.125, "weight": 0.25 }, "reader_interest": { "jev": 0.7, "judge_a": 0.72, "judge_b": null, "mean": 0.71, "spread": 0.02, "weight": 0.2 }, "audience_breadth": { "jev": 0.69, "judge_a": 0.67, "judge_b": null, "mean": 0.68, "spread": 0.02, "weight": 0.1 }, "practical_relevance": { "jev": 0.593, "judge_a": 0.5, "judge_b": null, "mean": 0.5463, "spread": 0.093, "weight": 0.15 }, "substantiated": { "jev": 0.68, "judge_a": 0.6, "judge_b": null, "mean": 0.64, "spread": 0.08, "weight": 0.1 }, "technical_depth": { "jev": 0.495, "judge_a": 0.3, "judge_b": null, "mean": 0.3975, "spread": 0.195, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.85, "judge_b": null, "mean": 0.88, "spread": 0.06, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.35, "judge_b": null, "mean": 0.205, "spread": 0.29, "weight": -0.15 }, "overclaim": { "jev": 0.28, "judge_a": 0.4, "judge_b": null, "mean": 0.34, "spread": 0.12, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 9, "front_page": false, "url": "https://news.ycombinator.com/item?id=50214701" }, "lobsters": null, "line": "Elsewhere: HN 9 pts, not on front page", "checked_at": "2026-10-24T08:54:10.558Z" } }, { "rank": 230, "id": "st_20261024_69", "title": "Attention people who think UB is fine", "url": "https://blog.habets.se/2026/10/Attention-people-who-think-UB-is-fine.html", "author": null, "site": "blog.habets.se", "published_at": "2026-10-23T02:58:35.000Z", "reading_minutes": 3, "depth": "two_judges", "format": "analysis_opinion", "summary": "A blog post by Mattias Habets demonstrates that undefined behavior in C is not harmlessly limited to a segfault. Using compiler output from GCC and clang, he shows that after a std::unreachable() call or a null pointer write, the compiler may emit code that runs unrelated functions, writes wrong values…", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "programming" ] }, "score": { "total": 0.5544, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.433, "judge_a": 0.3, "judge_b": null, "mean": 0.3662, "spread": 0.133, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.6, "judge_b": null, "mean": 0.57, "spread": 0.06, "weight": 0.2 }, "audience_breadth": { "jev": 0.607, "judge_a": 0.55, "judge_b": null, "mean": 0.5783, "spread": 0.057, "weight": 0.1 }, "practical_relevance": { "jev": 0.67, "judge_a": 0.6, "judge_b": null, "mean": 0.635, "spread": 0.07, "weight": 0.15 }, "substantiated": { "jev": 0.933, "judge_a": 0.8, "judge_b": null, "mean": 0.8667, "spread": 0.133, "weight": 0.1 }, "technical_depth": { "jev": 0.963, "judge_a": 0.85, "judge_b": null, "mean": 0.9063, "spread": 0.113, "weight": 0.1 }, "new_information": { "jev": 0.95, "judge_a": 0.35, "judge_b": null, "mean": 0.65, "spread": 0.6, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.03, "judge_b": null, "mean": 0.035, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.3, "judge_b": null, "mean": 0.275, "spread": 0.05, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 2, "front_page": false, "url": "https://news.ycombinator.com/item?id=50203996" }, "lobsters": null, "line": "Elsewhere: HN 2 pts, not on front page", "checked_at": "2026-10-24T08:54:10.605Z" } }, { "rank": 231, "id": "st_20261024_438", "title": "PHP Vulnerability Wrap-Up, (Fri, Oct 23rd)", "url": "https://isc.sans.edu/diary/33450", "author": "SANS Internet Storm Center", "site": "SANS Internet Storm Center", "published_at": "2026-10-23T13:43:20.000Z", "first_published_at": "2026-10-23T13:43:20.000Z", "reading_minutes": 2, "depth": "two_judges", "format": "analysis_opinion", "summary": "SANS Internet Storm Center's Johannes Ullrich wrote a wrap-up of a PHP vulnerability that prompted a temporary raise to Infocon Green alert status. Vendors largely patched in sync with PHP.net under coordinated disclosure, though one vendor leaked details early.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "not_a_project", "categories": [ "security", "it" ] }, "score": { "total": 0.5538, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.735, "judge_a": 0.45, "judge_b": null, "mean": 0.5925, "spread": 0.285, "weight": 0.25 }, "reader_interest": { "jev": 0.66, "judge_a": 0.6, "judge_b": null, "mean": 0.63, "spread": 0.06, "weight": 0.2 }, "audience_breadth": { "jev": 0.747, "judge_a": 0.7, "judge_b": null, "mean": 0.7233, "spread": 0.047, "weight": 0.1 }, "practical_relevance": { "jev": 0.815, "judge_a": 0.7, "judge_b": null, "mean": 0.7575, "spread": 0.115, "weight": 0.15 }, "substantiated": { "jev": 0.21, "judge_a": 0.4, "judge_b": null, "mean": 0.305, "spread": 0.19, "weight": 0.1 }, "technical_depth": { "jev": 0.545, "judge_a": 0.3, "judge_b": null, "mean": 0.4225, "spread": 0.245, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.35, "judge_b": null, "mean": 0.63, "spread": 0.56, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.05, "judge_b": null, "mean": 0.06, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.2, "judge_b": null, "mean": 0.22, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.905Z" } }, { "rank": 232, "id": "st_20261024_1558", "title": "Truthful by Omission: Click Thinning for LLM-Native Advertising", "url": "https://arxiv.org/abs/2610.30999", "author": "Junhua Liu, Zhicheng Liang, Zejun Zhang et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T09:40:16.000Z", "reading_minutes": 47, "depth": "two_judges", "format": "original_research", "summary": "Researchers propose \"click thinning,\" a mechanism for auctioning sponsored passages inside generative AI answers, where an early ad's clicks can cannibalize later ads and a higher bid can reorder competitors.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5538, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.595, "judge_a": 0.4, "judge_b": null, "mean": 0.4975, "spread": 0.195, "weight": 0.25 }, "reader_interest": { "jev": 0.45, "judge_a": 0.55, "judge_b": null, "mean": 0.5, "spread": 0.1, "weight": 0.2 }, "audience_breadth": { "jev": 0.48, "judge_a": 0.35, "judge_b": null, "mean": 0.415, "spread": 0.13, "weight": 0.1 }, "practical_relevance": { "jev": 0.54, "judge_a": 0.4, "judge_b": null, "mean": 0.47, "spread": 0.14, "weight": 0.15 }, "substantiated": { "jev": 0.92, "judge_a": 0.8, "judge_b": null, "mean": 0.86, "spread": 0.12, "weight": 0.1 }, "technical_depth": { "jev": 0.993, "judge_a": 0.9, "judge_b": null, "mean": 0.9463, "spread": 0.093, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": null, "mean": 0.855, "spread": 0.11, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": null, "mean": 0.055, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.29, "judge_a": 0.25, "judge_b": null, "mean": 0.27, "spread": 0.04, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:10.917Z" } }, { "rank": 233, "id": "st_20261024_1464", "title": "RIPPLE: Recovery Internalization through Process-Level Policy Learning and Effect-Aware Credit for Long-Horizon Agentic Coding", "url": "https://arxiv.org/abs/2610.31276", "author": "Jingxuan Xu, Zihang Xu, Yuechen Luo et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T13:45:33.000Z", "reading_minutes": 35, "depth": "two_judges", "format": "original_research", "summary": "RIPPLE is a training framework for long-horizon coding agents that targets the \"Critique-to-Action Handoff Gap,\" where an agent identifies a failure in a critique but keeps acting on the same incorrect path.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.5536, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.63, "judge_a": 0.35, "judge_b": null, "mean": 0.49, "spread": 0.28, "weight": 0.25 }, "reader_interest": { "jev": 0.6, "judge_a": 0.4, "judge_b": null, "mean": 0.5, "spread": 0.2, "weight": 0.2 }, "audience_breadth": { "jev": 0.563, "judge_a": 0.45, "judge_b": null, "mean": 0.5067, "spread": 0.113, "weight": 0.1 }, "practical_relevance": { "jev": 0.59, "judge_a": 0.5, "judge_b": null, "mean": 0.545, "spread": 0.09, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.6, "judge_b": null, "mean": 0.7767, "spread": 0.353, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.8, "judge_b": null, "mean": 0.885, "spread": 0.17, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": null, "mean": 0.85, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": null, "mean": 0.05, "spread": 0, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.35, "judge_b": null, "mean": 0.3, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.037Z" } }, { "rank": 234, "id": "st_20261024_323", "title": "A good teacher meets the student where they are", "url": "https://www.perplexity.ai/hub/blog/a-good-teacher-meets-the-student-where-they-are", "author": null, "site": "Perplexity AI", "published_at": "2026-10-23T09:56:11.000Z", "reading_minutes": 10, "depth": "two_judges", "format": "original_research", "summary": "JOLT (Joint On-policy Learning and Teaching) proposes training a single model as both a privileged-information teacher and an unprivileged student, so that the teacher's guidance is shaped to be useful to the student rather than merely correct.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5535, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.618, "judge_a": 0.4, "judge_b": null, "mean": 0.5088, "spread": 0.218, "weight": 0.25 }, "reader_interest": { "jev": 0.63, "judge_a": 0.55, "judge_b": null, "mean": 0.59, "spread": 0.08, "weight": 0.2 }, "audience_breadth": { "jev": 0.57, "judge_a": 0.4, "judge_b": null, "mean": 0.485, "spread": 0.17, "weight": 0.1 }, "practical_relevance": { "jev": 0.55, "judge_a": 0.4, "judge_b": null, "mean": 0.475, "spread": 0.15, "weight": 0.15 }, "substantiated": { "jev": 0.897, "judge_a": 0.6, "judge_b": null, "mean": 0.7483, "spread": 0.297, "weight": 0.1 }, "technical_depth": { "jev": 0.97, "judge_a": 0.85, "judge_b": null, "mean": 0.91, "spread": 0.12, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.8, "judge_b": null, "mean": 0.85, "spread": 0.1, "weight": 0.1 }, "promotional": { "jev": 0.09, "judge_a": 0.15, "judge_b": null, "mean": 0.12, "spread": 0.06, "weight": -0.15 }, "overclaim": { "jev": 0.24, "judge_a": 0.35, "judge_b": null, "mean": 0.295, "spread": 0.11, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206593" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:10.828Z" } }, { "rank": 235, "id": "st_20261024_1587", "title": "Automating Feature Flag Cleanup at Scale with Multi-Agent LLM Systems: An Industrial Experience at DoorDash", "url": "https://arxiv.org/abs/2610.31351", "author": "Atharv Chandratre, Jai Datar", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T14:32:23.000Z", "reading_minutes": 18, "depth": "two_judges", "format": "original_research", "summary": "DoorDash describes a multi-agent LLM system built on Google ADK that automates removal of stale feature flags (called Dynamic Values internally) across its codebases, where roughly 1,659 stale flags remain and manual cleanup takes one to two hours each.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "established_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.5527, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.628, "judge_a": 0.3, "judge_b": null, "mean": 0.4638, "spread": 0.327, "weight": 0.25 }, "reader_interest": { "jev": 0.75, "judge_a": 0.6, "judge_b": null, "mean": 0.675, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.36, "judge_a": 0.4, "judge_b": null, "mean": 0.38, "spread": 0.04, "weight": 0.1 }, "practical_relevance": { "jev": 0.855, "judge_a": 0.6, "judge_b": null, "mean": 0.7275, "spread": 0.255, "weight": 0.15 }, "substantiated": { "jev": 0.923, "judge_a": 0.5, "judge_b": null, "mean": 0.7117, "spread": 0.423, "weight": 0.1 }, "technical_depth": { "jev": 0.96, "judge_a": 0.6, "judge_b": null, "mean": 0.78, "spread": 0.36, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.7, "judge_b": null, "mean": 0.805, "spread": 0.21, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.15, "judge_b": null, "mean": 0.11, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.43, "judge_a": 0.35, "judge_b": null, "mean": 0.39, "spread": 0.08, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.115Z" } }, { "rank": 236, "id": "st_20261024_1481", "title": "What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models", "url": "https://arxiv.org/abs/2610.31576", "author": "Jinya Sakurai, Xun Xu, Jaehong Yoon", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T17:15:04.000Z", "reading_minutes": 45, "depth": "two_judges", "format": "original_research", "summary": "This study revisits learned unmasking policies for masked diffusion language models, which commit multiple tokens in parallel. Treating supervised policies as distillation of a decoding procedure, the authors separately vary what is labeled committable, the training states, and the policy inputs.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5525, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.543, "judge_a": 0.4, "judge_b": null, "mean": 0.4713, "spread": 0.142, "weight": 0.25 }, "reader_interest": { "jev": 0.54, "judge_a": 0.4, "judge_b": null, "mean": 0.47, "spread": 0.14, "weight": 0.2 }, "audience_breadth": { "jev": 0.46, "judge_a": 0.25, "judge_b": null, "mean": 0.355, "spread": 0.21, "weight": 0.1 }, "practical_relevance": { "jev": 0.628, "judge_a": 0.5, "judge_b": null, "mean": 0.5638, "spread": 0.127, "weight": 0.15 }, "substantiated": { "jev": 0.97, "judge_a": 0.8, "judge_b": null, "mean": 0.885, "spread": 0.17, "weight": 0.1 }, "technical_depth": { "jev": 0.968, "judge_a": 0.85, "judge_b": null, "mean": 0.9088, "spread": 0.118, "weight": 0.1 }, "new_information": { "jev": 0.91, "judge_a": 0.8, "judge_b": null, "mean": 0.855, "spread": 0.11, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": null, "mean": 0.045, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.25, "judge_b": null, "mean": 0.25, "spread": 0, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.230Z" } }, { "rank": 237, "id": "st_20261024_1775", "title": "FIGFLEX: A Mixed-Precision FP-INT GEMM Accelerator Co-Designed with Low-Bit LLM Quantization", "url": "https://arxiv.org/abs/2610.30480", "author": "Faraz Tahmasebi, Haocheng Xu, Andrew Trauger et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T21:59:47.000Z", "depth": "two_judges", "format": "original_research", "summary": "FIGFLEX is a hardware-software co-design for running low-bit LLMs, where weights are stored as integers and multiplied against floating-point activations. Its accelerator uses a precision-flexible radix-4 datapath so that throughput rises as weight precision falls, including an INT3 mode that uses otherwise idle hardware.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai", "hardware" ] }, "score": { "total": 0.5523, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.653, "judge_a": 0.45, "judge_b": null, "mean": 0.5513, "spread": 0.202, "weight": 0.25 }, "reader_interest": { "jev": 0.67, "judge_a": 0.6, "judge_b": null, "mean": 0.635, "spread": 0.07, "weight": 0.2 }, "audience_breadth": { "jev": 0.433, "judge_a": 0.4, "judge_b": null, "mean": 0.4167, "spread": 0.033, "weight": 0.1 }, "practical_relevance": { "jev": 0.573, "judge_a": 0.5, "judge_b": null, "mean": 0.5363, "spread": 0.073, "weight": 0.15 }, "substantiated": { "jev": 0.653, "judge_a": 0.6, "judge_b": null, "mean": 0.6267, "spread": 0.053, "weight": 0.1 }, "technical_depth": { "jev": 0.8, "judge_a": 0.85, "judge_b": null, "mean": 0.825, "spread": 0.05, "weight": 0.1 }, "new_information": { "jev": 0.94, "judge_a": 0.8, "judge_b": null, "mean": 0.87, "spread": 0.14, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.1, "judge_b": null, "mean": 0.085, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.37, "judge_a": 0.35, "judge_b": null, "mean": 0.36, "spread": 0.02, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.385Z" } }, { "rank": 238, "id": "st_20261024_1141", "title": "WhatsApp MCP – An MCP server for interacting with WhatsApp", "url": "https://github.com/sahilPadmani/whatsapp-mcp", "author": null, "site": "GitHub", "published_at": "2026-10-23T18:40:14.000Z", "reading_minutes": 5, "depth": "two_judges", "format": "new_projects", "summary": "A Show HN project, WhatsApp MCP, is an open-source Model Context Protocol server that lets AI agents such as Claude Code or Cursor read, search, send, and react to WhatsApp messages on one linked account. Phone numbers stay on the server, and data is AES-256-GCM encrypted.", "also_covered_by": [], "labels": { "content_type": "release", "project_stage": "new_side_project", "categories": [ "programming", "ai" ] }, "score": { "total": 0.5521, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.485, "judge_a": 0.3, "judge_b": null, "mean": 0.3925, "spread": 0.185, "weight": 0.25 }, "reader_interest": { "jev": 0.64, "judge_a": 0.55, "judge_b": null, "mean": 0.595, "spread": 0.09, "weight": 0.2 }, "audience_breadth": { "jev": 0.373, "judge_a": 0.4, "judge_b": null, "mean": 0.3867, "spread": 0.027, "weight": 0.1 }, "practical_relevance": { "jev": 0.85, "judge_a": 0.7, "judge_b": null, "mean": 0.775, "spread": 0.15, "weight": 0.15 }, "substantiated": { "jev": 0.773, "judge_a": 0.5, "judge_b": null, "mean": 0.6367, "spread": 0.273, "weight": 0.1 }, "technical_depth": { "jev": 0.833, "judge_a": 0.6, "judge_b": null, "mean": 0.7163, "spread": 0.233, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.8, "judge_b": null, "mean": 0.86, "spread": 0.12, "weight": 0.1 }, "promotional": { "jev": 0.07, "judge_a": 0.15, "judge_b": null, "mean": 0.11, "spread": 0.08, "weight": -0.15 }, "overclaim": { "jev": 0.13, "judge_a": 0.2, "judge_b": null, "mean": 0.165, "spread": 0.07, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 3, "front_page": false, "url": "https://news.ycombinator.com/item?id=50213531" }, "lobsters": null, "line": "Elsewhere: HN 3 pts, not on front page", "checked_at": "2026-10-24T08:54:11.168Z" } }, { "rank": 239, "id": "st_20261024_1769", "title": "EgoHand: Monocular 3D Hand Reconstruction from Egocentric Video", "url": "https://arxiv.org/abs/2610.30293", "author": "Chao-Yeh Chen, Tiffany Chen, Lin Duan et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-21T18:09:27.000Z", "depth": "two_judges", "format": "original_research", "summary": "EgoHand is a monocular model that reconstructs 3D hand poses from egocentric video, trained on a 14-million-image corpus mixing synthetic data, in-the-wild video, and multi-camera motion capture.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "product_release", "categories": [ "ai" ] }, "score": { "total": 0.5513, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.645, "judge_a": 0.4, "judge_b": null, "mean": 0.5225, "spread": 0.245, "weight": 0.25 }, "reader_interest": { "jev": 0.62, "judge_a": 0.5, "judge_b": null, "mean": 0.56, "spread": 0.12, "weight": 0.2 }, "audience_breadth": { "jev": 0.4, "judge_a": 0.3, "judge_b": null, "mean": 0.35, "spread": 0.1, "weight": 0.1 }, "practical_relevance": { "jev": 0.785, "judge_a": 0.55, "judge_b": null, "mean": 0.6675, "spread": 0.235, "weight": 0.15 }, "substantiated": { "jev": 0.823, "judge_a": 0.65, "judge_b": null, "mean": 0.7367, "spread": 0.173, "weight": 0.1 }, "technical_depth": { "jev": 0.728, "judge_a": 0.7, "judge_b": null, "mean": 0.7138, "spread": 0.028, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.9, "judge_b": null, "mean": 0.915, "spread": 0.03, "weight": 0.1 }, "promotional": { "jev": 0.12, "judge_a": 0.15, "judge_b": null, "mean": 0.135, "spread": 0.03, "weight": -0.15 }, "overclaim": { "jev": 0.27, "judge_a": 0.3, "judge_b": null, "mean": 0.285, "spread": 0.03, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.442Z" } }, { "rank": 240, "id": "st_20261024_762", "title": "Claude Mods: Flipping the Harness Inside Out", "url": "https://latent.space/p/claude-mods/", "author": null, "site": "Latent Space", "published_at": "2026-10-23T15:28:11.000Z", "reading_minutes": 9, "depth": "three_judges", "format": "analysis_opinion", "summary": "Hrih News reports on Claude Mods, a plugin system for Claude Code that lets users write TypeScript or JavaScript functions which hook into the agent's events and change how prompts, tool calls, and interface rendering are handled.", "also_covered_by": [], "labels": { "content_type": "analysis_opinion", "project_stage": "product_release", "categories": [ "programming", "ai", "product" ] }, "score": { "total": 0.5507, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.578, "judge_a": 0.35, "judge_b": 0.62, "mean": 0.5158, "spread": 0.27, "weight": 0.25 }, "reader_interest": { "jev": 0.72, "judge_a": 0.6, "judge_b": 0.86, "mean": 0.7267, "spread": 0.26, "weight": 0.2 }, "audience_breadth": { "jev": 0.453, "judge_a": 0.5, "judge_b": 0.55, "mean": 0.5011, "spread": 0.097, "weight": 0.1 }, "practical_relevance": { "jev": 0.84, "judge_a": 0.6, "judge_b": 0.78, "mean": 0.74, "spread": 0.24, "weight": 0.15 }, "substantiated": { "jev": 0.57, "judge_a": 0.5, "judge_b": 0.68, "mean": 0.5833, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.773, "judge_a": 0.6, "judge_b": 0.67, "mean": 0.6808, "spread": 0.173, "weight": 0.1 }, "new_information": { "jev": 0.92, "judge_a": 0.6, "judge_b": 0.95, "mean": 0.8233, "spread": 0.35, "weight": 0.1 }, "promotional": { "jev": 0.12, "judge_a": 0.35, "judge_b": 0.28, "mean": 0.25, "spread": 0.23, "weight": -0.15 }, "overclaim": { "jev": 0.34, "judge_a": 0.4, "judge_b": 0.38, "mean": 0.3733, "spread": 0.06, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50210675" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:11.292Z" } }, { "rank": 241, "id": "st_20261024_1386", "title": "Accelerating Wildfire Hazard Mapping Across Canada with an AI Surrogate Modelling Framework", "url": "https://arxiv.org/abs/2610.30170", "author": "Hager Radi Abdelwahed, Toluwaleke Olutayo, Chirag Goel et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-16T19:06:07.000Z", "reading_minutes": 52, "depth": "two_judges", "format": "original_research", "summary": "A Mila and Natural Resources Canada paper presents a multi-task U-Net-style AI surrogate that emulates BurnP3+ wildfire simulations, predicting burn probability, fire intensity, and rate of spread across Canadian regions.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5506, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.573, "judge_a": 0.35, "judge_b": null, "mean": 0.4613, "spread": 0.223, "weight": 0.25 }, "reader_interest": { "jev": 0.55, "judge_a": 0.55, "judge_b": null, "mean": 0.55, "spread": 0, "weight": 0.2 }, "audience_breadth": { "jev": 0.377, "judge_a": 0.3, "judge_b": null, "mean": 0.3383, "spread": 0.077, "weight": 0.1 }, "practical_relevance": { "jev": 0.66, "judge_a": 0.4, "judge_b": null, "mean": 0.53, "spread": 0.26, "weight": 0.15 }, "substantiated": { "jev": 0.98, "judge_a": 0.8, "judge_b": null, "mean": 0.89, "spread": 0.18, "weight": 0.1 }, "technical_depth": { "jev": 0.96, "judge_a": 0.8, "judge_b": null, "mean": 0.88, "spread": 0.16, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.85, "judge_b": null, "mean": 0.875, "spread": 0.05, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.05, "judge_b": null, "mean": 0.055, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.34, "judge_a": 0.25, "judge_b": null, "mean": 0.295, "spread": 0.09, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.556Z" } }, { "rank": 242, "id": "st_20261024_397", "title": "I get 215 GitHub notifications a day, so I built a bot to triage them", "url": "https://paolino.me/copilot-triage/", "author": "Carmine Paolino", "site": "Carmine Paolino", "published_at": "2026-10-23T10:47:28.000Z", "reading_minutes": 8, "depth": "two_judges", "format": "engineering_writeup", "summary": "Copilot Triage is an open source bot by the maintainer of RubyLLM, built after GitHub's compiled Agentic Workflows produced 2,035 lines of YAML and noisy self-reported failures.", "also_covered_by": [], "labels": { "content_type": "engineering_writeup", "project_stage": "product_release", "categories": [ "programming", "ai" ] }, "score": { "total": 0.5505, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.483, "judge_a": 0.3, "judge_b": null, "mean": 0.3913, "spread": 0.183, "weight": 0.25 }, "reader_interest": { "jev": 0.75, "judge_a": 0.6, "judge_b": null, "mean": 0.675, "spread": 0.15, "weight": 0.2 }, "audience_breadth": { "jev": 0.553, "judge_a": 0.4, "judge_b": null, "mean": 0.4767, "spread": 0.153, "weight": 0.1 }, "practical_relevance": { "jev": 0.92, "judge_a": 0.6, "judge_b": null, "mean": 0.76, "spread": 0.32, "weight": 0.15 }, "substantiated": { "jev": 0.963, "judge_a": 0.6, "judge_b": null, "mean": 0.7817, "spread": 0.363, "weight": 0.1 }, "technical_depth": { "jev": 0.943, "judge_a": 0.6, "judge_b": null, "mean": 0.7713, "spread": 0.343, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.6, "judge_b": null, "mean": 0.75, "spread": 0.3, "weight": 0.1 }, "promotional": { "jev": 0.08, "judge_a": 0.45, "judge_b": null, "mean": 0.265, "spread": 0.37, "weight": -0.15 }, "overclaim": { "jev": 0.16, "judge_a": 0.3, "judge_b": null, "mean": 0.23, "spread": 0.14, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50206956" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:11.422Z" } }, { "rank": 243, "id": "st_20261024_485", "title": "How Much Damage Do Fake Shops Cause?", "url": "https://nebty-id.com/en/doppelcart-fake-shop-economics/", "author": null, "site": "nebty", "published_at": "2026-10-23T11:00:54.000Z", "reading_minutes": 12, "depth": "two_judges", "format": "original_research", "summary": "Nebty's investigation into the DoppelCart fake-shop network, which it linked to 119,000 domains, estimates that fake shops could cause $1.2 billion to $5.8 billion in annual losses worldwide. The network's checkout collects card details and bank confirmation codes, which can enable card fraud at $100 to $500 per victim.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "security" ] }, "score": { "total": 0.5505, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.573, "judge_a": 0.6, "judge_b": null, "mean": 0.5862, "spread": 0.027, "weight": 0.25 }, "reader_interest": { "jev": 0.53, "judge_a": 0.7, "judge_b": null, "mean": 0.615, "spread": 0.17, "weight": 0.2 }, "audience_breadth": { "jev": 0.473, "judge_a": 0.55, "judge_b": null, "mean": 0.5117, "spread": 0.077, "weight": 0.1 }, "practical_relevance": { "jev": 0.5, "judge_a": 0.45, "judge_b": null, "mean": 0.475, "spread": 0.05, "weight": 0.15 }, "substantiated": { "jev": 0.953, "judge_a": 0.55, "judge_b": null, "mean": 0.7517, "spread": 0.403, "weight": 0.1 }, "technical_depth": { "jev": 0.733, "judge_a": 0.5, "judge_b": null, "mean": 0.6163, "spread": 0.233, "weight": 0.1 }, "new_information": { "jev": 0.93, "judge_a": 0.75, "judge_b": null, "mean": 0.84, "spread": 0.18, "weight": 0.1 }, "promotional": { "jev": 0.06, "judge_a": 0.08, "judge_b": null, "mean": 0.07, "spread": 0.02, "weight": -0.15 }, "overclaim": { "jev": 0.19, "judge_a": 0.5, "judge_b": null, "mean": 0.345, "spread": 0.31, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": { "points": 1, "front_page": false, "url": "https://news.ycombinator.com/item?id=50207068" }, "lobsters": null, "line": "Elsewhere: HN 1 pts, not on front page", "checked_at": "2026-10-24T08:54:11.509Z" } }, { "rank": 244, "id": "st_20261024_1423", "title": "ProQuant: Preserving Expert Quality over Quantity in Ultra-Low-Bit Quantization of Mixture-of-Experts Models", "url": "https://arxiv.org/abs/2610.30779", "author": "Jiaqi Chen, Xianglong Yan, Chengzhu Bao et al.", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T06:01:37.000Z", "reading_minutes": 35, "depth": "two_judges", "format": "original_research", "summary": "ProQuant is a weight-only post-training quantization framework for Mixture-of-Experts language models that targets very low bit widths. It combines Prune-or-Prioritize Mixed-Precision Allocation, which jointly decides expert pruning and bit-width assignment, with Routing-Consistent RMSNorm Finetune, which compensates quantization error without shifting routing behavior.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5504, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.608, "judge_a": 0.35, "judge_b": null, "mean": 0.4788, "spread": 0.258, "weight": 0.25 }, "reader_interest": { "jev": 0.57, "judge_a": 0.4, "judge_b": null, "mean": 0.485, "spread": 0.17, "weight": 0.2 }, "audience_breadth": { "jev": 0.5, "judge_a": 0.3, "judge_b": null, "mean": 0.4, "spread": 0.2, "weight": 0.1 }, "practical_relevance": { "jev": 0.71, "judge_a": 0.55, "judge_b": null, "mean": 0.63, "spread": 0.16, "weight": 0.15 }, "substantiated": { "jev": 0.973, "judge_a": 0.72, "judge_b": null, "mean": 0.8467, "spread": 0.253, "weight": 0.1 }, "technical_depth": { "jev": 0.955, "judge_a": 0.8, "judge_b": null, "mean": 0.8775, "spread": 0.155, "weight": 0.1 }, "new_information": { "jev": 0.9, "judge_a": 0.7, "judge_b": null, "mean": 0.8, "spread": 0.2, "weight": 0.1 }, "promotional": { "jev": 0.05, "judge_a": 0.05, "judge_b": null, "mean": 0.05, "spread": 0, "weight": -0.15 }, "overclaim": { "jev": 0.31, "judge_a": 0.3, "judge_b": null, "mean": 0.305, "spread": 0.01, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.701Z" } }, { "rank": 245, "id": "st_20261024_1426", "title": "What Lies Beyond Perception: A Partition Theory of the Limits of AI Evaluation and Oversight", "url": "https://arxiv.org/abs/2610.30803", "author": "Ahmed M. Mahran", "site": "arXiv", "published_at": "2026-10-23T04:00:00.000Z", "first_published_at": "2026-10-22T06:28:29.000Z", "reading_minutes": 78, "depth": "two_judges", "format": "original_research", "summary": "A partition theory of AI evaluation argues that an evaluator can certify only what its own perceptual resolution allows. Under this framing, a passing score shows only that a model discriminates at least as finely as the evaluator, not that it lacks any ability.", "also_covered_by": [], "labels": { "content_type": "original_research", "project_stage": "not_a_project", "categories": [ "ai" ] }, "score": { "total": 0.5504, "contested": false, "flag_penalty": 0, "breakdown": { "significance": { "jev": 0.7, "judge_a": 0.45, "judge_b": null, "mean": 0.575, "spread": 0.25, "weight": 0.25 }, "reader_interest": { "jev": 0.57, "judge_a": 0.6, "judge_b": null, "mean": 0.585, "spread": 0.03, "weight": 0.2 }, "audience_breadth": { "jev": 0.64, "judge_a": 0.5, "judge_b": null, "mean": 0.57, "spread": 0.14, "weight": 0.1 }, "practical_relevance": { "jev": 0.553, "judge_a": 0.45, "judge_b": null, "mean": 0.5013, "spread": 0.102, "weight": 0.15 }, "substantiated": { "jev": 0.447, "judge_a": 0.6, "judge_b": null, "mean": 0.5233, "spread": 0.153, "weight": 0.1 }, "technical_depth": { "jev": 0.948, "judge_a": 0.7, "judge_b": null, "mean": 0.8238, "spread": 0.248, "weight": 0.1 }, "new_information": { "jev": 0.89, "judge_a": 0.6, "judge_b": null, "mean": 0.745, "spread": 0.29, "weight": 0.1 }, "promotional": { "jev": 0.04, "judge_a": 0.05, "judge_b": null, "mean": 0.045, "spread": 0.01, "weight": -0.15 }, "overclaim": { "jev": 0.25, "judge_a": 0.35, "judge_b": null, "mean": 0.3, "spread": 0.1, "weight": -0.15 } } }, "claims": [], "elsewhere": { "hn": null, "lobsters": null, "line": "Elsewhere: HN —", "checked_at": "2026-10-24T08:54:11.717Z" } } ] }