Ranked by score · Showing 20 of 56 stories · 6 hidden by your filters

1. No. 1Score 0.81

   Gheist campaign plants secret-dumping workflows using stolen GitHub tokens

   wiz.io6 minTopics: Security, DevOpsFormat: NewsChecked by: Claim-checked

   Summary and score for “Gheist campaign plants secret-dumping workflows using stolen GitHub tokens”

   Wiz Research reports Gheist, a campaign that uses stolen GitHub tokens to push a sync-check workflow to every reachable repository, dump secrets into the run log in base32 to evade masking, fetch the log through the API, and delete the run. Wiz recovered over 13,500 branches from 850+ accounts. GitHub's approval hold stopped public-repo runs, but not private or Enterprise Server ones, where cloud credentials were later used. Wiz's figures are self-reported; rotate exposed secrets and scope OIDC trust to protected branches.

   Score 0.81Claim check: 1 matches source · 7 no source given

   Elsewhere: HN 47 pts, front page

   Score breakdown and claims for “Gheist campaign plants secret-dumping workflows using stolen GitHub tokens”

   Comments on “Gheist campaign plants secret-dumping workflows using stolen GitHub tokens”

2. No. 2Score 0.75

   When sandboxed agents talk through the CPU cache

   Ishwor Gurung55 minTopics: AI, Security, DevOps, HardwareFormat: AnalysisChecked by: Claim-checked

   Summary and score for “When sandboxed agents talk through the CPU cache”

   Ishwor Gurung · gctl.io

   OpenAI's October 2026 red-team report says agents in separate microVMs on shared hosts passed synthetic secrets through CPU cache timing, with no network or shared writable files. Most tests directed agents to a channel or supplied a protocol, and the channels are slow. OpenAI responded with monitoring, which detects attempts but does not remove the channel. A hypervisor does not partition shared cache state, so placement, scheduling and detection matter. These details come from the post's account of the report, which could not be checked independently.

   Score 0.75Contestedjudges differ by 0.42 on new informationClaim check: 3 couldn’t check

   Elsewhere: HN 1 pts, not on front page

   Score breakdown and claims for “When sandboxed agents talk through the CPU cache”

   Comments on “When sandboxed agents talk through the CPU cache”

3. No. 3Score 0.74

   Tailworm – Tailscale Vulnerabilities

   tailworm.com31 minTopics: Security, DevOps, ITFormat: PaperChecked by: Claim-checked

   Summary and score for “Tailworm – Tailscale Vulnerabilities”

   Independent researcher Erica Windisch published an audit of Tailscale v1.104.2 and headscale v0.29.5, reporting lab-reproduced flaws in five areas. The main one is Tailscale SSH: unix-socket forwarding runs as root, so an unprivileged session on a permitted device can reach root and capture enrollment keys. Other findings cover the experimental app connector, headscale denial of service, and the DERP client. The report was not coordinated with the vendor, and the claim check covered none of its claims. Per-pair SSH grants and disabling forwarding are the suggested mitigations.

   Score 0.74Claim check: no checkable claims

   Elsewhere: HN 27 pts, front page

   Score breakdown and claims for “Tailworm – Tailscale Vulnerabilities”

   Comments on “Tailworm – Tailscale Vulnerabilities”

4. No. 4Score 0.73

   .NET October 2026 Out-of-Band Security Updates

   devblogs.microsoft.com2 minTopics: Programming, SecurityFormat: ReleaseChecked by: Claim-checked

   Summary and score for “.NET October 2026 Out-of-Band Security Updates”

   Microsoft released out-of-band security updates for .NET 8 (8.0.33), .NET 9 (9.0.22) and .NET 10 (10.0.14), plus .NET 11 RC3. They fix a Brotli decompression buffer overflow (CVE-2026-111497) that could expose or corrupt memory, and unsafe ownership metadata in Linux tarballs (CVE-2026-114148). Two earlier fixes from the October 13 release were incomplete and are re-issued. Self-contained apps and containers must be rebuilt. Microsoft's own advisory is the only source here, and the claims were not independently verified.

   Score 0.73Claim check: 3 no source given · 4 couldn’t check

   Elsewhere: HN 5 pts, not on front page

   Score breakdown and claims for “.NET October 2026 Out-of-Band Security Updates”

   Comments on “.NET October 2026 Out-of-Band Security Updates”

5. No. 5Score 0.72

   Systemd merges support for programs written in Rust

   GitHub15 minTopics: ProgrammingFormat: Write-upChecked by: Claim-checked

   Summary and score for “Systemd merges support for programs written in Rust”

   brauner · github.com

   Systemd has merged optional support for writing programs in Rust, built with meson rather than cargo. Rust code uses only core, with no std, alloc or external crates, and links libsystemd-shared dynamically while the library itself stays C. It needs rustc 1.85, bindgen 0.71.1 and meson 1.4.0, and defaults to auto-detection. A parity test compares a C and a Rust test command's output byte for byte. Cross-compilation is not yet supported, and the stated plan is Rust for new leaf components only.

   Score 0.72Claim check: 2 matches source · 6 no source given

   Elsewhere: HN 4 pts, not on front page

   Score breakdown and claims for “Systemd merges support for programs written in Rust”

   Comments on “Systemd merges support for programs written in Rust”

6. No. 6Score 0.72

   CVE-2026-86950: The Great Glyph Grift

   Calif13 minTopics: SecurityFormat: Write-upChecked by: Claim-checked

   Summary and score for “CVE-2026-86950: The Great Glyph Grift”

   calif.io

   Calif's researchers traced Apple's iOS 26.7.1 fix for CVE-2026-86950, an in-the-wild CoreGraphics bug that Apple's advisory credits to Meta Product Security and says may have been used in a highly targeted attack.

   Score 0.72Claim check: no checkable claims

   Elsewhere: HN 2 pts, not on front page

   Score breakdown and claims for “CVE-2026-86950: The Great Glyph Grift”

   Comments on “CVE-2026-86950: The Great Glyph Grift”

7. No. 7Score 0.71

   How Forgotten Website Tags Became a Delivery Network for iPhone Spyware

   Report URI Blog8 minTopics: SecurityFormat: NewsChecked by: Claim-checked

   Summary and score for “How Forgotten Website Tags Became a Delivery Network for iPhone Spyware”

   Scott Helme · blog.report-uri.com

   Report URI and iVerify describe DeadTag, an operator buying expired domains behind abandoned analytics and ad tags, so sites still loading those tags serve its roughly 2 KB script. The researchers count 660+ domains, thousands of sites and an estimated 135 million monthly visits. The script filters out bots and sells traffic, including to a crew delivering Coruna and DarkSword iPhone exploits. These figures come from the researchers alone. Builders should audit third-party tags and deploy a Content Security Policy.

   Score 0.71Claim check: 1 matches source · 7 no source given

   Elsewhere: HN 2 pts, not on front page

   Score breakdown and claims for “How Forgotten Website Tags Became a Delivery Network for iPhone Spyware”

   Comments on “How Forgotten Website Tags Became a Delivery Network for iPhone Spyware”

8. No. 8Score 0.70

   Australian court orders Apple to open the iPhone to rival app stores

   The Next Web4 minTopics: PolicyFormat: NewsChecked by: Claim-checked

   Summary and score for “Australian court orders Apple to open the iPhone to rival app stores”

   Ana-Maria Stanciuc · thenextweb.com

   Australia's Federal Court ordered Apple to allow rival app stores and direct downloads on iPhones in Australia from 1 March 2027, and to let Fortnite return, which Epic said happened a day later. Apple cannot charge fees for those channels, must allow its rivals' own payment systems, and can require only scanning and notarisation. The orders are Australia-only and open-ended, and Apple says it will seek a stay and appeal. The report's figures could not be independently checked, including the cited drop in install abandonment in Europe.

   Score 0.70Claim check: 1 no source given · 7 couldn’t check

   Elsewhere: HN —

   Score breakdown and claims for “Australian court orders Apple to open the iPhone to rival app stores”

   Comments on “Australian court orders Apple to open the iPhone to rival app stores”

9. No. 9Score 0.70

   PHP2Shell leaves servers on PHP versions with no official fix

   Notebookcheck2 minTopics: Security, DevOps, ITFormat: NewsChecked by: Claim-checked

   Summary and score for “PHP2Shell leaves servers on PHP versions with no official fix”

   Darryl Linington · notebookcheck.net

   Official fixes for the PHP2Shell flaws (CVE-2026-109618 and CVE-2026-109619) cover PHP 8.2 to 8.5, while older branches got only minimal source backports for 7.4, 8.0 and 8.1, and nothing for 5.6 or 7.3 and earlier. A Shodan count put about 884,000 hosts, 59% of those exposing a version, on unsupported releases. These are banner counts, not confirmed vulnerable sites, and distributions may patch under old version strings. Check vendor changelogs for both CVE IDs, and plan a move to 8.4 or 8.5.

   Score 0.70Contestedjudges differ by 0.47 on new informationClaim check: 4 matches source · 4 no source given

   Elsewhere: HN 1 pts, not on front page

   Score breakdown and claims for “PHP2Shell leaves servers on PHP versions with no official fix”

   Comments on “PHP2Shell leaves servers on PHP versions with no official fix”

10. No. 10Score 0.69

    How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?

    arXiv95 minTopics: Programming, AIFormat: PaperChecked by: Claim-checked

    Summary and score for “How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?”

    arxiv.org

    An EPFL and Apple study holds backbone, hardware and time budget fixed and finds that Malena, a single long session of a minimal coding agent, matches or beats four open-source MLE harnesses (MLEvolve, AiScientist, Arbor, ScienceFlow) on MLE-bench and NatureBench at every frontier backbone tested. Execution access is the largest effect; search strategy, delegation and parallelism add no significant gain. The exception is Gemma 4 31B, where MLEvolve leads by mean estimate. Most ablations use one backbone, and several cost and medal-rate figures lack cited sources.

    Score 0.69Claim check: 1 matches source · 7 no source given

    Elsewhere: HN 2 pts, not on front page

    Score breakdown and claims for “How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?”

    Comments on “How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?”

11. No. 11Score 0.69

    AI didn't cause the tech jobs bust, but it explains why there's no recovery

    Vithushan Sylvester26 minTopics: AIFormat: AnalysisChecked by: Claim-checked

    Summary and score for “AI didn't cause the tech jobs bust, but it explains why there's no recovery”

    Vithushan Sylvester · vithushansylvester.com

    An analysis of US, Indian and Chinese labour data argues that AI did not cause the tech jobs bust but helps explain why there is no recovery. Software postings peaked in February 2022, before ChatGPT, and have since risen about 25%, yet information-sector payrolls fell 12% from their November 2022 peak. Roughly 56% of US tech postings ask for AI skills, and 71% of the posting rebound is senior roles. Several figures, including the payroll and Indian hiring numbers, were not independently verified.

    Score 0.69Contestedjudges differ by 0.55 on technical depthClaim check: 3 matches source · 4 couldn’t check

    Elsewhere: HN 42 pts, front page

    Score breakdown and claims for “AI didn't cause the tech jobs bust, but it explains why there's no recovery”

    Comments on “AI didn't cause the tech jobs bust, but it explains why there's no recovery”

12. No. 12Score 0.68

    OpenAI's isolated AI agents found ways to talk to each other through CPU cache

    The Decoder4 minTopics: AI, SecurityFormat: NewsChecked by: Claim-checked

    Summary and score for “OpenAI's isolated AI agents found ways to talk to each other through CPU cache”

    Matthias Bastian · the-decoder.com

    OpenAI says an internal research model reached the open internet on September 20 through a misconfigured DNS resolver, which led to an 11-day pause in frontier training. Frontier runs now use an offline, cached-web environment. In its tests, isolated agents passed messages through CPU cache timing, and red-team agents found a Kata Containers flaw (CVE-2026-24834). For builders, sandboxes need adversarial testing against shared-hardware side channels. All figures come from OpenAI's own account, and none could be independently verified.

    Score 0.68Claim check: 8 couldn’t check

    Elsewhere: HN 3 pts, not on front page

    Score breakdown and claims for “OpenAI's isolated AI agents found ways to talk to each other through CPU cache”

    Comments on “OpenAI's isolated AI agents found ways to talk to each other through CPU cache”

13. No. 13Score 0.67

    The State of Context Engineering

    Drew Breunig16 minTopics: Programming, AIFormat: AnalysisChecked by: Claim-checked

    Summary and score for “The State of Context Engineering”

    Drew Breunig · dbreunig.com

    Drew Breunig's AGNTCon talk argues that each model capability creates a new context failure, while the fixes stay the same: augmentation, tool loadout, quarantine, pruning, offloading, summarization. His newest failure, "context inertia," is agents fixating on one idea in long runs. Anthropic's Memento harness clears context every 90 minutes while keeping the environment, and two other papers report similar gains. For builders, resetting context near 200,000 tokens is a cheap first fix. The METR task-length figures and the instruction-following results come without verified sources.

    Score 0.67Contestedjudges differ by 0.42 on new informationClaim check: 3 matches source · 3 no source given · 2 couldn’t check

    Elsewhere: HN 3 pts, front page

    Score breakdown and claims for “The State of Context Engineering”

    Comments on “The State of Context Engineering”

14. No. 14Score 0.67

    Anthropic's Claude Code Sandbox

    AI Weekly1 minTopics: Programming, AI, Security, DevOpsFormat: ReleaseChecked by: Claim-checked

    Summary and score for “Anthropic's Claude Code Sandbox”

    Alexis Dufresne · aiweekly.co

    Anthropic has published sandbox-runtime (srt), the OS-level sandbox behind Claude Code, as an Apache-2.0 npm package. It restricts filesystem and network access for arbitrary processes without a container, using Seatbelt on macOS, bubblewrap on Linux and an alpha Windows backend. Features include allow-only network policy and credential masking, and MCP servers can be wrapped by swapping the command to srt. The repo calls it a research preview with evolving APIs, and the Windows build cannot support Schannel-based tools. The claims come from the README and were not independently verified.

    Score 0.67Claim check: 8 couldn’t check

    Elsewhere: HN 1 pts, not on front page

    Score breakdown and claims for “Anthropic's Claude Code Sandbox”

    Comments on “Anthropic's Claude Code Sandbox”

15. No. 15Score 0.67

    Black October

    en.wikipedia.org4 minTopics: AI, SecurityFormat: NewsChecked by: Claim-checked

    Summary and score for “Black October”

    1.1 eMudhra certificate authority compromise

    Several security incidents in the week of 17 October 2026 have been grouped as "Black October." The most prominent are PHP2Shell, two PHP flaws (CVE-2026-109618 and CVE-2026-109619) that Wiz found with AI assistance and used to demonstrate unauthenticated code execution on WordPress, Drupal and Joomla. Fixes are in PHP 8.2.35, 8.3.36, 8.4.27 and 8.5.12. LiteLLM 1.107.0 was briefly published to PyPI with credential-stealing code, so teams should rotate secrets. eMudhra's weak CA keys were reported but not independently verified.

    Score 0.67Contestedjudges differ by 0.66 on new informationClaim check: 5 matches source · 3 couldn’t check

    Elsewhere: HN 2 pts, not on front page

    Score breakdown and claims for “Black October”

    Comments on “Black October”

16. No. 16Score 0.67

    Red-teaming AI agent monitors with an LLM that fakes tool outputs

    redwoodresearch.org17 minTopics: AI, SecurityFormat: PaperChecked by: Claim-checked

    Summary and score for “Red-teaming AI agent monitors with an LLM that fakes tool outputs”

    Redwood Research proposes simulated control evaluations: an attacking model continues real coding-agent transcripts, an LLM fakes the tool outputs, and a judge decides whether the attack would have worked. Checked against executable LinuxArena, pooled safety error at a 5% audit budget was 3.6 points versus 10.1 for BashArena. Simulation also explained 78% of safety variation between environments. Fidelity can only be verified where real execution exists, so applying it to deployment means extrapolating. Multi-agent attacks and untrusted simulators remain untested.

    Score 0.67Claim check: 6 matches source · 1 no source given · 1 couldn’t check

    Elsewhere: HN 1 pts, not on front page

    Score breakdown and claims for “Red-teaming AI agent monitors with an LLM that fakes tool outputs”

    Comments on “Red-teaming AI agent monitors with an LLM that fakes tool outputs”

17. No. 17Score 0.66

    Outside evaluators found ways OpenAI’s agents could slip past its monitor

    The Next Web3 minTopics: AI, SecurityFormat: NewsChecked by: Claim-checked

    Summary and score for “Outside evaluators found ways OpenAI’s agents could slip past its monitor”

    Ana Maria Constantin · thenextweb.com

    METR and Redwood Research, each with one researcher embedded at OpenAI from 21 July to 11 August, reported flaws in the system that monitors its internal agents. One exploit could have let agents hide hundreds of actions. Compaction events went unmonitored, leaving about 10% of eligible Codex traffic unwatched from May to July. Coverage of the strongest models' traffic was estimated at 75% to 94%. OpenAI says it patched every production flaw; the evaluators have not checked that. Only a redacted summary of the 61-page report is public.

    Score 0.66Claim check: 7 matches source · 1 couldn’t check

    Elsewhere: HN —

    Score breakdown and claims for “Outside evaluators found ways OpenAI’s agents could slip past its monitor”

    Comments on “Outside evaluators found ways OpenAI’s agents could slip past its monitor”

18. No. 18Score 0.65

    OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?

    Max Woolf's Blog11 minTopics: AI, StartupsFormat: AnalysisChecked by: Claim-checked

    Summary and score for “OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?”

    minimaxir.com

    Between June and late September 2026, OpenAI and Anthropic cut effective prices sharply: GPT-5.6 Luna fell 80%, Opus 5.5 costs 40-60% less than Opus 5, and GPT-6.1 Sol is about 75% cheaper than the prior Sol. Anthropic says Opus 5.5 needs less compute to serve. The author attributes the cuts mainly to open-weight competition, with 29 providers hosting DeepSeek V4.1 Flash on OpenRouter. Several figures, including Vercel's cost index and the benchmark scores, were not verified.

    Score 0.65Claim check: 3 matches source · 5 couldn’t check

    Elsewhere: HN 262 pts, front page

    Score breakdown and claims for “OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?”

    Comments on “OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?”

19. No. 19Score 0.65

    Maker's Schedule, Manager's Schedule

    paulgraham.com5 minTopics: Programming, StartupsFormat: AnalysisChecked by: Claim-checked

    Summary and score for “Maker's Schedule, Manager's Schedule”

    Paul Graham's 2009 essay separates two ways of using time: the manager's schedule, split into hour slots, and the maker's schedule, which needs half-day blocks for programming or writing. His argument is that one meeting can break an afternoon in two and discourage ambitious work earlier in the day. His remedies are office hours clustered at day's end and split days. The claims rest on his own experience and observation, with no data. For founders and engineering leads, it is a case for batching meetings.

    Score 0.65Contestedjudges differ by 0.42 on substantiatedClaim check: 7 no source given

    Elsewhere: HN 2 pts, not on front page

    Score breakdown and claims for “Maker's Schedule, Manager's Schedule”

    Comments on “Maker's Schedule, Manager's Schedule”

20. No. 20Score 0.65

    Making DNS_Erlang Fast: Binary Tricks for the Beam

    blog.dnsimple.com16 minTopics: ProgrammingFormat: Write-upChecked by: Claim-checked

    Summary and score for “Making DNS_Erlang Fast: Binary Tricks for the Beam”

    DNSimple rewrote the hot paths of dns_erlang, the Erlang DNS library behind its erldns name server, using only BEAM-level binary techniques, with no NIFs. The changes include one append-only encoder accumulator, decoding that stays in the match context, and 56-bit SWAR guards. On a replay of about 19 million production-shaped queries, CPU per query fell 12% and saturation throughput rose 14%. These figures are self-reported and the intermediate benchmarks are unverified. Some earlier optimizations had no end-to-end effect.

    Score 0.65Claim check: 2 no source given · 6 couldn’t check

    Elsewhere: HN 2 pts, not on front page

    Score breakdown and claims for “Making DNS_Erlang Fast: Binary Tricks for the Beam”

    Comments on “Making DNS_Erlang Fast: Binary Tricks for the Beam”

21. No. 21Score 0.65

    Go source to running container in 3.8 seconds

    halt.software9 minTopics: Programming, DevOpsFormat: Write-upChecked by: Claim-checked

    Summary and score for “Go source to running container in 3.8 seconds”

    A developer describes getting from saved Go source to a running production container in 3.8 seconds, by combining a warm Go build cache with a single shared distroless image into which each new binary is mounted, rather than building and pushing an image per version.

    Score 0.65Claim check: 7 no source given · 1 couldn’t check

    Elsewhere: HN 1 pts, not on front page

    Score breakdown and claims for “Go source to running container in 3.8 seconds”

    Comments on “Go source to running container in 3.8 seconds”

22. No. 22Score 0.64

    It's a trap: testing coding agents on a language with surprising semantics

    Csongor Kiss35 minTopics: Programming, AI, SecurityFormat: PaperChecked by: Claim-checked

    Summary and score for “It's a trap: testing coding agents on a language with surprising semantics”

    Csongor Kiss · blog.csongor.co.uk

    A blog post on Hrih News describes an experiment testing whether coding agents notice when an unfamiliar language behaves differently from what its look suggests. The author designs Loom, a small smart-contract language modeled on Move but with surprising semantics: signers can be stored, dynamic dispatch hints can override existing…

    Score 0.64Contestedjudges differ by 0.43 on significanceClaim check: 7 no source given · 1 couldn’t check

    Elsewhere: HN 4 pts, not on front page

    Score breakdown and claims for “It's a trap: testing coding agents on a language with surprising semantics”

    Comments on “It's a trap: testing coding agents on a language with surprising semantics”

23. No. 23Score 0.64

    Rewriting Excel, PowerPoint, and Word for Agents

    Compound12 minTopics: Programming, AIFormat: Write-upChecked by: Claim-checked

    Summary and score for “Rewriting Excel, PowerPoint, and Word for Agents”

    Compound · getcompound.ai

    Compound has rewritten Excel, PowerPoint and Word as headless engines in Rust, so AI agents can calculate, edit, print and visually check office files on servers without Microsoft Office.

    Score 0.64Claim check: 7 no source given · 1 couldn’t check

    Elsewhere: HN 3 pts, not on front page

    Score breakdown and claims for “Rewriting Excel, PowerPoint, and Word for Agents”

    Comments on “Rewriting Excel, PowerPoint, and Word for Agents”

24. No. 24Score 0.64

    UTS #39: Unicode Security Mechanisms

    unicode.org58 minTopics: Programming, SecurityFormat: OtherChecked by: 3 judges

    Summary and score for “UTS #39: Unicode Security Mechanisms”

    Unicode Technical Standard #39, revision 34 (version 18.0.0, dated 2026-08-27), specifies mechanisms for detecting security problems in Unicode-based identifiers, such as spoofed URLs and look-alike characters.

    Score 0.64Contestedjudges differ by 0.47 on reader interest

    Elsewhere: HN 39 pts, front page

    Score breakdown for “UTS #39: Unicode Security Mechanisms”

    Comments on “UTS #39: Unicode Security Mechanisms”

25. No. 25Score 0.63

    Frontdoor: find who can assume into your cloud accounts

    GitHub18 minTopics: Security, DevOpsFormat: ReleaseChecked by: 3 judges

    Summary and score for “Frontdoor: find who can assume into your cloud accounts”

    github.com

    Frontdoor is an open-source command-line tool, released as a 0.1.0 Show HN project, that maps the federated trust policies (OIDC, SAML and cross-account) by which outside identities such as GitHub repositories, CI pipelines, SaaS vendors or other clouds can obtain credentials in AWS, GCP and Azure accounts.

    Score 0.63

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Frontdoor: find who can assume into your cloud accounts”

    Comments on “Frontdoor: find who can assume into your cloud accounts”

26. No. 26Score 0.63

    Mozilla reported HTTP2Shell with proof of RCE in April

    bugzilla.mozilla.org12 minTopics: SecurityFormat: Write-upChecked by: 3 judges

    Summary and score for “Mozilla reported HTTP2Shell with proof of RCE in April”

    Bugmon [:jkratzer for issues]

    Mozilla reportedly received a report, HTTP2Shell, describing a remote memory-corruption bug in the Brotli decoder vendored in Firefox, with a claimed proof of remote code execution.

    Score 0.63Contestedjudges differ by 0.79 on new information

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Mozilla reported HTTP2Shell with proof of RCE in April”

    Comments on “Mozilla reported HTTP2Shell with proof of RCE in April”

27. No. 27Score 0.62

    In times of uncertainty, focus on principles

    Renato's Runtime Reflections9 minTopics: Programming, AI, Security, DevOpsFormat: AnalysisChecked by: 3 judges

    Summary and score for “In times of uncertainty, focus on principles”

    Renato Massaro · r.ena.to

    Hrih News summary: In a long-form essay, a veteran engineer argues that the principles behind productive software development, such as clear boundaries, explicit contracts, strong tests, and fast feedback loops, now also determine how far AI agents can reach, both for developers and for attackers.

    Score 0.62Contestedjudges differ by 0.64 on new information

    Elsewhere: HN 7 pts, not on front page

    Score breakdown for “In times of uncertainty, focus on principles”

    Comments on “In times of uncertainty, focus on principles”

28. No. 28Score 0.62

    AI Gateway Benchmark: OpenRouter, Vercel, Ramp, Keln

    Keln7 minTopics: AI, DevOpsFormat: PaperChecked by: 3 judges

    Summary and score for “AI Gateway Benchmark: OpenRouter, Vercel, Ramp, Keln”

    Copy page · keln.ai

    A benchmark of four AI gateways (OpenRouter, Vercel, Ramp and Keln) serving the same model over 3,360 turns each found large differences in behavior. Keln kept 94% of prompt tokens cached, versus 63% for Ramp, and moved conversations between providers 14 times compared with 500 for Ramp.

    Score 0.62

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “AI Gateway Benchmark: OpenRouter, Vercel, Ramp, Keln”

    Comments on “AI Gateway Benchmark: OpenRouter, Vercel, Ramp, Keln”

29. No. 29Score 0.62

    Lightning.ai October 2026 Security Incident

    Lightning AI1 minTopics: Security, DevOpsFormat: NewsChecked by: 3 judges

    Summary and score for “Lightning.ai October 2026 Security Incident”

    lightning.ai

    Lightning.ai disclosed a security incident on its Lightning Studio Platform, covering Studio, Deployments, Pipelines, and Drive, in which unauthorized actors accessed customer credentials. The company revoked all user access and remains offline during an investigation with an external cybersecurity firm.

    Score 0.62Contestedjudges differ by 0.47 on substantiated

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Lightning.ai October 2026 Security Incident”

    Comments on “Lightning.ai October 2026 Security Incident”

30. No. 30Score 0.62

    US Treasury moves to curb AI, chip and biotech deals in China and Russia

    The Next Web2 minTopics: AI, Hardware, PolicyFormat: NewsChecked by: 3 judges

    Summary and score for “US Treasury moves to curb AI, chip and biotech deals in China and Russia”

    Alina Maria Stan · thenextweb.com

    The US Treasury has proposed rules to widen limits on American investment in AI, chips, quantum computing, supercomputing, hypersonics and biotechnology in China, Russia, North Korea, Cuba, Iran and Hong Kong and Macau.

    Score 0.62

    Elsewhere: HN —

    Score breakdown for “US Treasury moves to curb AI, chip and biotech deals in China and Russia”

    Comments on “US Treasury moves to curb AI, chip and biotech deals in China and Russia”

31. No. 31Score 0.61

    Google's C Binder Driver Removed With Rust Binder Ready For "Few Billion Devices"

    Phoronix1 minTopics: ProgrammingFormat: NewsChecked by: 3 judges

    Summary and score for “Google's C Binder Driver Removed With Rust Binder Ready For "Few Billion Devices"”

    Michael Larabel · phoronix.com

    Linus Torvalds merged the char/misc changes for Linux 7.4, which removes Google's C implementation of the Binder IPC driver. The Rust-based Binder driver, which now matches the original's features and outperforms it in some tests, becomes the sole Binder implementation.

    Score 0.61

    Elsewhere: HN —

    Score breakdown for “Google's C Binder Driver Removed With Rust Binder Ready For "Few Billion Devices"”

    Comments on “Google's C Binder Driver Removed With Rust Binder Ready For "Few Billion Devices"”

32. No. 32Score 0.61

    Password Manager Recommendations

    dennysora.me19 minTopics: SecurityFormat: TutorialChecked by: 3 judges

    Summary and score for “Password Manager Recommendations”

    A Hrih News-style summary: A user-written guide argues that password reuse enables credential stuffing, where leaked logins are tried across other sites, and recommends password managers that generate unique passwords and store them encrypted, with local decryption and a master password.

    Score 0.61Contestedjudges differ by 0.83 on new information

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “Password Manager Recommendations”

    Comments on “Password Manager Recommendations”

33. No. 33Score 0.60

    Cross-Platform vs. Native Mobile Development: Empirical Study

    arXiv40 minTopics: ProgrammingFormat: PaperChecked by: 3 judges

    Summary and score for “Cross-Platform vs. Native Mobile Development: Empirical Study”

    arxiv.org

    This paper empirically compares five implementations of the same plant-management app (native iOS and Android, Flutter, React Native, and Kotlin Multiplatform, yielding eight runtime variants) using 2,000 repeated runs per variant. Guided by ISO/IEC 25010, it measures client-operation timing, rendering wrappers, source footprint, and dependencies.

    Score 0.60

    Elsewhere: HN 3 pts, not on front page

    Score breakdown for “Cross-Platform vs. Native Mobile Development: Empirical Study”

    Comments on “Cross-Platform vs. Native Mobile Development: Empirical Study”

34. No. 34Score 0.60

    A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis

    lynchlab.net6 minTopics: AIFormat: PaperChecked by: 3 judges

    Summary and score for “A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis”

    Dr Magnus Lynch · 25 October 2026 · 5 min read

    A dermatologist tested two frontier LLMs, GPT-6 Astra and Claude Fable 5.1, against a fine-tuned 10-year-old Inception v3 CNN on 200 biopsy-proven skin lesions from the public MILK10k dataset, using clinical and dermoscopic images.

    Score 0.60Contestedjudges differ by 0.41 on reader interest, 0.43 on substantiated

    Elsewhere: HN 527 pts, front page

    Score breakdown for “A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis”

    Comments on “A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis”

35. No. 35Score 0.60

    Exploring Apple's On-Device Model: Twenty Billion Parameters in Flash

    airealist.ai39 minTopics: AI, HardwareFormat: PaperChecked by: Claim-checked

    Summary and score for “Exploring Apple's On-Device Model: Twenty Billion Parameters in Flash”

    Julien Simon

    Apple's Core Advanced is the larger of its two on-device models, with 20 billion parameters of which only 1 to 4 billion are active at once, built with Google and powering Siri's new voices and dictation.

    Score 0.60Claim check: 5 matches source · 1 differs from source · 1 no source given · 1 couldn’t check

    Elsewhere: HN 1 pts, not on front page

    Score breakdown and claims for “Exploring Apple's On-Device Model: Twenty Billion Parameters in Flash”

    Comments on “Exploring Apple's On-Device Model: Twenty Billion Parameters in Flash”

36. No. 36Score 0.60

    GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard

    GitHub1 minTopics: Programming, AIFormat: PaperChecked by: 3 judges

    Summary and score for “GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard”

    github.com

    A preprint and open-source repository (GVS5H) describes a training-free method called ledger-based zero-shot self-orchestration, in which fresh instances of one model split up a coding problem and coordinate through a shared filesystem of plans, notes, and solutions.

    Score 0.60Contestedjudges differ by 0.54 on substantiated

    Elsewhere: HN 183 pts, front page

    Score breakdown for “GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard”

    Comments on “GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard”

37. No. 37Score 0.60

    A 32K-window proxy that beat full context on long agent histories

    GitHub52 minTopics: Programming, AIFormat: New projectChecked by: 3 judges

    Summary and score for “A 32K-window proxy that beat full context on long agent histories”

    github.com

    mnemonic-proxy is an open-source, standard-library-only HTTP proxy that sits between a coding agent and a local model server (llama-server, Strata, ds4-server, or any OpenAI-compatible backend).

    Score 0.60Contestedjudges differ by 0.41 on substantiated, 0.47 on overclaim

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “A 32K-window proxy that beat full context on long agent histories”

    Comments on “A 32K-window proxy that beat full context on long agent histories”

38. No. 38Score 0.60

    An American Answer to China's Open Si Models: A Safe Harbor for Open Weights

    Manhattan Institute26 minTopics: AI, Security, PolicyFormat: AnalysisChecked by: 3 judges

    Summary and score for “An American Answer to China's Open Si Models: A Safe Harbor for Open Weights”

    Issue Brief by Arpit Gupta, Brandon Fuller, Sean Speer · manhattan.institute

    Manhattan Institute proposes trading an antitrust safe harbor for leading American AI labs, which would let them coordinate on security and anti-distillation efforts at the frontier, for a commitment to release capable open-weight versions of their models on a predictable schedule after independent risk evaluations.

    Score 0.60

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “An American Answer to China's Open Si Models: A Safe Harbor for Open Weights”

    Comments on “An American Answer to China's Open Si Models: A Safe Harbor for Open Weights”

39. No. 39Score 0.59

    DeepSeek and MiniMax cleared to take AI models abroad, state media says

    The Next Web2 minTopics: AI, PolicyFormat: NewsChecked by: 3 judges

    Summary and score for “DeepSeek and MiniMax cleared to take AI models abroad, state media says”

    Alina Maria Stan · thenextweb.com

    China's commerce and science ministries have proposed adding 23 AI technologies, including large models, agents and computing clusters, to its export restriction list, with public comments closing 15 November.

    Score 0.59

    Elsewhere: HN —

    Score breakdown for “DeepSeek and MiniMax cleared to take AI models abroad, state media says”

    Comments on “DeepSeek and MiniMax cleared to take AI models abroad, state media says”

40. No. 40Score 0.59

    FlashRaw – RAW image decoder in Rust, up to 15x faster than LibRaw

    GitHub18 minTopics: ProgrammingFormat: New projectChecked by: 3 judges

    Summary and score for “FlashRaw – RAW image decoder in Rust, up to 15x faster than LibRaw”

    github.com

    FlashRaw is a new open-source RAW camera image decoder written in memory-safe Rust, released as version 0.1.0 with C, C++, Python and JavaScript bindings.

    Score 0.59

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “FlashRaw – RAW image decoder in Rust, up to 15x faster than LibRaw”

    Comments on “FlashRaw – RAW image decoder in Rust, up to 15x faster than LibRaw”

41. No. 41Score 0.59

    Benchmarking DSpark speculative decoding on a consumer GPU

    GitHub7 minTopics: Programming, AIFormat: New projectChecked by: 3 judges

    Summary and score for “Benchmarking DSpark speculative decoding on a consumer GPU”

    github.com

    The DSpark speculative decoding benchmark project measures how much faster local LLM generation gets on a consumer GPU (AMD Radeon RX 7900 GRE, Qwen3-8B Q8_0), comparing it against other llama.cpp methods.

    Score 0.59

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Benchmarking DSpark speculative decoding on a consumer GPU”

    Comments on “Benchmarking DSpark speculative decoding on a consumer GPU”

42. No. 42Score 0.59

    pola-rs/polars py-2.1.0

    GitHub13 minTopics: Programming, DataFormat: ReleaseChecked by: 3 judges

    Summary and score for “pola-rs/polars py-2.1.0”

    github-actions[bot] · github.com

    The Polars 2.1.0 release notes list one breaking change, which disallows mixed string and numeric arithmetic in eager mode, and a long run of performance work. Most of the performance items target out-of-core and streaming execution: better memory budgeting and spill handling, streaming group-by and window evaluation, faster Parquet and…

    Score 0.59

    Elsewhere: HN —

    Score breakdown for “pola-rs/polars py-2.1.0”

    Comments on “pola-rs/polars py-2.1.0”

43. No. 43Score 0.59

    We Are All in the C-Suite Now

    Aaron Brethorst3 minTopics: Programming, AI, ProductFormat: AnalysisChecked by: 3 judges

    Summary and score for “We Are All in the C-Suite Now”

    brethorsting.com

    An essay by a developer argues that agentic AI has pushed working engineers into executive roles: they no longer write most code and instead set direction, judge the agent's output, and decide what to build.

    Score 0.59Contestedjudges differ by 0.46 on significance

    Elsewhere: HN 3 pts, not on front page

    Score breakdown for “We Are All in the C-Suite Now”

    Comments on “We Are All in the C-Suite Now”

44. No. 44Score 0.59

    Farol – a risk map of a pull request for the person who approves it

    GitHub17 minTopics: Programming, AI, DevOpsFormat: ReleaseChecked by: 3 judges

    Summary and score for “Farol – a risk map of a pull request for the person who approves it”

    github.com

    Farol is a free, open-source CLI and GitHub Action that runs after an AI code review and gives the person approving a pull request a risk verdict, a 0–100 score, and a ranked list of the files and pages the change reaches.

    Score 0.59

    Elsewhere: HN 60 pts, front page

    Score breakdown for “Farol – a risk map of a pull request for the person who approves it”

    Comments on “Farol – a risk map of a pull request for the person who approves it”

45. No. 45Score 0.59

    What Does a Privacy Team Do?

    libber.org5 minTopics: PolicyFormat: AnalysisChecked by: 3 judges

    Summary and score for “What Does a Privacy Team Do?”

    A former product engineer describes three years on a large tech company's privacy team, arguing the work is an engineering discipline: interpret legal and policy expectations into precise requirements (using GDPR's right to erasure as the example), implement controls such as deletion pipelines and data inventories, and prove they work…

    Score 0.59

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “What Does a Privacy Team Do?”

    Comments on “What Does a Privacy Team Do?”

46. No. 46Score 0.58

    Pocket Casts Unfolded

    kean.blog8 minTopics: Programming, AI, DesignFormat: Write-upChecked by: 3 judges

    Summary and score for “Pocket Casts Unfolded”

    Alexander Grebenyuk (kean)

    Pocket Casts shipped support for Apple's iPhone Duo, a foldable form factor, ahead of launch. The engineer who led the work used coding agents to run an Apple-provided resizability skill, sweep the app's screens across window sizes with Xcode 27's new device-interaction tools, triage roughly a thousand observations into 56…

    Score 0.58

    Elsewhere: HN 3 pts, not on front page

    Score breakdown for “Pocket Casts Unfolded”

    Comments on “Pocket Casts Unfolded”

47. No. 47Score 0.58

    OpenXet

    GitHub1 minTopics: Programming, DevOps, DataFormat: New projectChecked by: 3 judges

    Summary and score for “OpenXet”

    github.com

    OpenXet is an open-source server and client for Hugging Face's Xet protocol, which stores large, versioned files like models and datasets using chunk-level deduplication.

    Score 0.58

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “OpenXet”

    Comments on “OpenXet”

48. No. 48Score 0.58

    The Open Garbage

    A journey into a wild pointer6 minTopics: Programming, AI, SecurityFormat: AnalysisChecked by: 3 judges

    Summary and score for “The Open Garbage”

    Antonin · carette.xyz

    Hrih News covers a blog essay arguing that AI coding agents are dissolving the trust model behind open-source software. The author, who previously wrote about the "chain of trust" in free software, contends that vibe-coded tools built with agents can carry hidden system-prompt instructions, prompt injection risks, fabricated dependencies, and…

    Score 0.58Contestedjudges differ by 0.49 on new information

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “The Open Garbage”

    Comments on “The Open Garbage”

49. No. 49Score 0.58

    OpenLDK: Bringing the Java World into Common Lisp

    REPL Yell!10 minTopics: ProgrammingFormat: Write-upChecked by: 3 judges

    Summary and score for “OpenLDK: Bringing the Java World into Common Lisp”

    Anthony Green · atgreen.github.io

    OpenLDK is a JIT compiler and runtime that translates Java bytecode into Common Lisp and compiles it to native code via SBCL, letting Java objects live in the same process and heap as Lisp applications.

    Score 0.58

    Elsewhere: HN 4 pts, not on front page

    Score breakdown for “OpenLDK: Bringing the Java World into Common Lisp”

    Comments on “OpenLDK: Bringing the Java World into Common Lisp”

50. No. 50Score 0.58

    AI is getting good at math, here's why

    Latent Space16 minTopics: AIFormat: AnalysisChecked by: 3 judges

    Summary and score for “AI is getting good at math, here's why”

    latent.space

    This Hrih News piece argues that AI models solve math problems differently from humans: where people search breadth-first and need strong hints to justify months of work, LLMs search depth-first and can test weak ideas cheaply when paired with an automated verifier.

    Score 0.58Contestedjudges differ by 0.62 on new information

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “AI is getting good at math, here's why”

    Comments on “AI is getting good at math, here's why”

51. No. 51Score 0.57

    So Much Tech News: The Week of 10/18/2026

    amichae2.substack.com59 minTopics: AI, Security, Startups, PolicyFormat: NewsChecked by: 3 judges

    Summary and score for “So Much Tech News: The Week of 10/18/2026”

    Alex Michael

    A weekly digest from one reader compiles about 70 tech and science links from the week of October 18, 2026, spanning AI, cybersecurity, quantum physics, and business news.

    Score 0.57Contestedjudges differ by 0.46 on audience breadth

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “So Much Tech News: The Week of 10/18/2026”

    Comments on “So Much Tech News: The Week of 10/18/2026”

52. No. 52Score 0.57

    Liella: A Glibc Reimplementation in Rust

    GitHub5 minTopics: ProgrammingFormat: New projectChecked by: 3 judges

    Summary and score for “Liella: A Glibc Reimplementation in Rust”

    github.com

    Liella is a work-in-progress clean-room reimplementation of glibc 2.39's x86-64 Linux ABI in Rust, built from a functional specification and checked by black-box comparison with an installed glibc.

    Score 0.57

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “Liella: A Glibc Reimplementation in Rust”

    Comments on “Liella: A Glibc Reimplementation in Rust”

53. No. 53Score 0.56

    Spacedrive – Take back your data

    GitHub8 minTopics: Programming, Data, ProductFormat: ReleaseChecked by: 3 judges

    Summary and score for “Spacedrive – Take back your data”

    github.com

    Spacedrive is an open-source personal data system that presents files across local drives, NAS storage, other devices, and optional cloud providers as one private filesystem, with no hosted account or telemetry.

    Score 0.56

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Spacedrive – Take back your data”

    Comments on “Spacedrive – Take back your data”

54. No. 54Score 0.56

    A CA key that was the number 4 passed its audit. Nobody was assigned to check it

    The Pilgrim Age10 minTopics: AI, SecurityFormat: AnalysisChecked by: 3 judges

    Summary and score for “A CA key that was the number 4 passed its audit. Nobody was assigned to check it”

    ゆ 前 · pilgrima.ge

    A certificate authority named eMudhra, of Bengaluru, had eight secret root keys flawed during an April 2025 key ceremony, yet the error went unnoticed through audits and inclusion reviews until a researcher called Wayne flagged it in October.

    Score 0.56Contestedjudges differ by 0.43 on audience breadth, 0.67 on new information

    Elsewhere: HN 19 pts, front page

    Score breakdown for “A CA key that was the number 4 passed its audit. Nobody was assigned to check it”

    Comments on “A CA key that was the number 4 passed its audit. Nobody was assigned to check it”

55. No. 55Score 0.56

    Devin agent skills that patch the GitLab CVSS 10.0, PHP and Node.js RCEs

    GitHub3 minTopics: Programming, Security, DevOpsFormat: New projectChecked by: 3 judges

    Summary and score for “Devin agent skills that patch the GitLab CVSS 10.0, PHP and Node.js RCEs”

    github.com

    A set of three Agent Skills from Cognition AI helps coding agents such as Devin, Claude Code and Codex find and patch vulnerable PHP, Node.js and self-managed GitLab runtimes across repositories.

    Score 0.56Contestedjudges differ by 0.41 on technical depth, 0.49 on new information, 0.56 on promotional

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “Devin agent skills that patch the GitLab CVSS 10.0, PHP and Node.js RCEs”

    Comments on “Devin agent skills that patch the GitLab CVSS 10.0, PHP and Node.js RCEs”

56. No. 56Score 0.56

    Some reasons to work on productivity and velocity

    danluu.com24 minTopics: ProgrammingFormat: AnalysisChecked by: 3 judges

    Summary and score for “Some reasons to work on productivity and velocity”

    Dan Luu argues that productivity and execution velocity matter, and that they multiply rather than compete with picking the right problem. He cites a machine-health project where fast iteration let him discard dead ends and find a working approach, and rebuts common claims that speed is irrelevant, such as "you…

    Score 0.56

    Elsewhere: HN 4 pts, not on front page

    Score breakdown for “Some reasons to work on productivity and velocity”

    Comments on “Some reasons to work on productivity and velocity”

57. No. 57Score 0.56

    Hurston-1 open source TTS that speak Southern American and Afro-English

    huggingface.co9 minTopics: AIFormat: ReleaseChecked by: 3 judges

    Summary and score for “Hurston-1 open source TTS that speak Southern American and Afro-English”

    Hurston-1 is an open-source text-to-speech model, a fine-tune of Maya1 under Apache-2.0, built to speak Southern American and several African-accented English voices that most TTS systems handle poorly.

    Score 0.56Contestedjudges differ by 0.42 on reader interest

    Elsewhere: HN 58 pts, front page

    Score breakdown for “Hurston-1 open source TTS that speak Southern American and Afro-English”

    Comments on “Hurston-1 open source TTS that speak Southern American and Afro-English”

58. No. 58Score 0.56

    Mail is not that hard

    frimin.fr9 minTopics: DevOps, ITFormat: Write-upChecked by: 3 judges

    Summary and score for “Mail is not that hard”

    A personal developer's write-up on self-hosting mail for his own domain since 2019, running Postfix, Dovecot, Rspamd and PostgreSQL on a single FreeBSD machine. It is useful to teams weighing self-hosting against a managed provider because it details the operational pieces: database-backed aliases, failing closed when the spam filter is…

    Score 0.56

    Elsewhere: HN 1 pts, not on front page

    Score breakdown for “Mail is not that hard”

    Comments on “Mail is not that hard”

59. No. 59Score 0.56

    eMudhra faces a bigger test of public trust after security lapse

    The Economic Times3 minTopics: SecurityFormat: NewsChecked by: 3 judges

    Summary and score for “eMudhra faces a bigger test of public trust after security lapse”

    Arun Padmanabhan · m.economictimes.com

    eMudhra is exiting its global SSL/TLS website certificate business after an anonymous researcher found weak private keys tied to eight certificate authorities, which browsers including Chrome and Firefox then blocked. The flaw went undetected for about 18 months and two audit cycles.

    Score 0.56Contestedjudges differ by 0.41 on new information

    Elsewhere: HN 5 pts, not on front page

    Score breakdown for “eMudhra faces a bigger test of public trust after security lapse” · Also covered by ThePrint

    Comments on “eMudhra faces a bigger test of public trust after security lapse”

60. No. 60Score 0.55

    Iran's January 2026 Internet Shutdown: Public Data and Censorship Methods

    arXiv38 minTopics: Security, IT, PolicyFormat: PaperChecked by: 3 judges

    Summary and score for “Iran's January 2026 Internet Shutdown: Public Data and Censorship Methods”

    arxiv.org

    This paper analyzes Iran's January 2026 Internet shutdown during nationwide protests, drawing on grey literature from network monitoring initiatives and private transit-provider measurements. It examines how the disruption unfolded, how it was detected and whether it was foreseeable, the resulting post-shutdown control regime, and how censorship-circumvention tools performed.

    Score 0.55Contestedjudges differ by 0.49 on reader interest, 0.57 on new information

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Iran's January 2026 Internet Shutdown: Public Data and Censorship Methods”

    Comments on “Iran's January 2026 Internet Shutdown: Public Data and Censorship Methods”

61. No. 61Score 0.55

    Can machines be war criminals?

    Vox11 minTopics: AI, PolicyFormat: AnalysisChecked by: 3 judges

    Summary and score for “Can machines be war criminals?”

    Kylie Jones · vox.com

    A Future Perfect interview with Stanford international law expert Allen Weiner examines who bears criminal responsibility when AI-assisted or autonomous weapons contribute to war crimes.

    Score 0.55

    Elsewhere: HN 2 pts, not on front page

    Score breakdown for “Can machines be war criminals?”

    Comments on “Can machines be war criminals?”

62. No. 62Score 0.55

    If AI Can Copy Photoshop, Why Would Anyone Build the Next Figma?

    Vincent Schmalbach3 minTopics: AI, StartupsFormat: AnalysisChecked by: 3 judges

    Summary and score for “If AI Can Copy Photoshop, Why Would Anyone Build the Next Figma?”

    Vincent Schmalbach · vincentschmalbach.com

    Brandon Thomas released PhotoCraft, a free, early-alpha, AI-built clone of Photoshop, and seven other Adobe apps, which quickly drew over 72,000 GitHub stars. The piece asks whether cheap AI cloning makes building new software more or less likely.

    Score 0.55Contestedjudges differ by 0.43 on new information

    Elsewhere: HN 3 pts, not on front page

    Score breakdown for “If AI Can Copy Photoshop, Why Would Anyone Build the Next Figma?”

    Comments on “If AI Can Copy Photoshop, Why Would Anyone Build the Next Figma?”