The week in tech, ranked blind:16 to 22 November 2026

Every story from this week’s editions, in one list by score. The score comes from the AI judges, who read each story’s text with the site and the author hidden. Each title leads to the story’s page, with every answer they gave.

Week 47 of 20261 edition60 storiesRubric v2.3Updated each morning until the Sunday edition

Editions: Mon 16

The top twenty

Ranked by score · 20 of 60 stories

  1. No. 1Score 0.79

    NoisyTCP: Linux Kernel TCP Pre-Auth Remote Corruption (CVE-2026-99288)

    Judge 2 Major remote kernel TCP vulnerability affecting nearly all Linux servers, with a clear technical description, mitigations, and practical impact for builders. Limited reproduction detail and a stated lab-only exploit temper substantiation.

  2. No. 2Score 0.76

    Skyfall Attack: Microarchitectural Data Sampling on AMD

    Judge 2 Substantive academic security research with detailed microarchitectural technical content and evidence; highly significant for cloud, confidential computing and CPU security, though audience is specialized rather than all of tech.

  3. No. 3Score 0.73

    AWS made an official local emulator

    Judge 2 Hands-on evaluation of AWS's new local emulator with specific, reproducible findings on gaps, CI strategy, and alternatives. Substantive and practical for builders, though the subject is a single vendor tool with a moderate audience.

  4. No. 5Score 0.72

    Found 380 eMudhra-Linked Root Certificates on My Laptop

    Judge 2 Concrete, evidence-backed security finding (scan of 9M files, forged-cert acceptance tests) affecting many developer tools with actionable remediation; scope limited to eMudhra roots and a self-described personal scanner tool.

  5. No. 6Score 0.72

    PyPI code that breaks on Python 3.12 and 3.13

    Judge 2 Data-driven analysis of PyPI imports of modules removed in Python 3.12/3.13, with careful methodology and practical fixes; original research with strong evidence, though the topic is niche and modestly significant.

  6. No. 7Score 0.70

    Slack's Migration to a Cellular Architecture

    Judge 2 Detailed engineering write-up of a cellular AZ-drain architecture with Envoy weighted clusters and siloing; concrete, practically useful for infrastructure builders, with reasonable but company-internal evidence.

  7. No. 8Score 0.69

    Sandboxes for AI Agents

    Judge 2 Practical, well-organized guide to sandboxing for coding agents with a taxonomy of designs and tradeoffs. Largely opinion and synthesis with some cited examples; little original data, so substantiation is moderate.}

  8. No. 9Score 0.68

    LLM Releases, Token Limits, and Pricing Since GPT 3.5

    Judge 2 Data-driven analysis of LLM release cadence, context and output limits, and pricing trends, useful to builders. Sourcing is mostly embedded charts, and some claims are interpretive guesses, so substantiation is moderate.

  9. No. 10Score 0.68

    GPU says 35% busy. Busy doing what?

    Judge 2 Detailed, measured engineering write-up on LLM inference GPU utilisation with reproducible flags and numbers; practical for builders, narrow audience, modest significance, some self-promotion of their tool.

  10. No. 11Score 0.67

    Berlin’s massive data leak: How to dissect it

    Judge 2 Major government ransomware leak with a practitioner interview on analysis methods and tooling; technical detail is moderate, sourcing is an interview with an operator, and policy angle is strong.

  11. No. 12Score 0.67

    No Free Lunch, No Free Rollback: Where Agents Need Engineers

    Judge 2 Opinion essay on AI coding agents and irreversible vs reversible mistakes; relevant to builders, argued with cited sources but mostly conceptual thesis, with a few cited studies and some overreach in its framing.

  12. No. 13Score 0.67

    Musings on long-context fine-tuning

    Judge 2 Detailed, reproducible engineering write-up on long-context LoRA fine-tuning with FSDP, memory tuning, and shared-prefix packing for hybrid attention/recurrent models; highly technical and practical for ML builders, but niche in audience.}

  13. No. 14Score 0.66

    Old Bitly Links Are Being Hijacked

    Judge 2 Well-substantiated security investigation of Bitly back-half link hijacking with verifiable methodology and data; strong practical relevance for builders managing short links, though audience is somewhat niche and the article is long and update-heavy.

  14. No. 15Score 0.66

    Editing one row of an LLM's output layer to stop it overthinking

    Judge 2 Strong technical depth with exact math and reproducible code on an active LLM topic, but small sample sizes, single runs, and self-reported benchmarks limit substantiation and breadth of impact.

  15. No. 16Score 0.66

    I let LLMs play a 1988 detective gamebook 646 times

    Judge 2 A blinded, well-substantiated experiment on LLM agents learning from retries, with clear practical lessons for agent builders, though the subject is a toy gamebook and the findings are narrow.

  16. No. 17Score 0.66

    Misaligned AI behaviors that indicate loss-of-control risk

    Judge 2 A taxonomy of AI loss-of-control behaviors, a framework document rather than a study. Highly relevant to AI builders and safety; substantiated by examples but not empirical evidence, with high novelty versus the recent list.

  17. No. 18Score 0.66

    Building a Serverless ClickHouse

    Judge 2 Detailed, well-evidenced engineering writeup on building serverless ClickHouse with compute/storage separation, routing, and cost measurements; a product launch with real technical depth, of interest to data and infra builders but narrow in audience.

  18. No. 19Score 0.64

    Static recompiler that runs Windows x64 on Apple Silicon

    Judge 2 Deep, well-documented technical project with measured benchmarks and reproducible steps; clearly limited scope and honest caveats. Niche audience but substantive, novel engineering content.

  19. No. 20Score 0.64

    Rust and Git worktrees: share one Cargo build

    Judge 2 Detailed, well-evidenced engineering write-up on Cargo build sharing across git worktrees with reproducible measurements; useful to Rust builders but narrow in audience and not a landmark.

The top five in each topic

Programming & dev tools

The top 5 of 33 stories

AI & machine learning

The top 5 of 30 stories

Security

The top 5 of 11 stories

Infrastructure & DevOps

The top 5 of 12 stories

IT & networking

5 stories

  1. No. 4Score 0.62

    Code Red: Y2K Is Here

Data

The top 5 of 6 stories

Hardware & devices

The top 5 of 9 stories

Product

1 story

Design & UX

1 story

Fintech

1 story

Policy & regulation

4 stories

  1. No. 2Score 0.62

    Code Red: Y2K Is Here

A story the pipeline marked as a close repeat of one listed above it appears once here; its day’s edition lists both. A story can belong to several topics. How ranking works