What's actually new
The week's most-read essays, ranked by what they add beyond what AI already knows and what's already online. Week of 26 October 2026.
-
1
Qwenputer: a fully functioning, deterministic RISC-V CPU inside Qwen3.5 divagr.com · Original Thinker92verdict
A substantial original build that compiles a working RV32IM CPU into Qwen3.5 weights. It backs the claim with its own validation and benchmark numbers, though readers cannot check them here.
-
2
Oldest GUI application that still runs on Windows 11 blog.enlyth.com · Original Thinker90verdict
A hands-on probe of Windows 11 backward compatibility that diagnoses why 1992 NT pre-release binaries do or don't run.
-
3
Agents in Games labs.ycombinator.com · Original Thinker90verdict
Original multi-agent game experiments with concrete logs showing arbitrary hierarchies, compounding caution and contrasting GPT and Claude swarm styles, though sample sizes are small.
-
4
The Goose Programming Language · Goose aardappel.github.io · Original Thinker88verdict
An original language design writeup with concrete mechanisms and candid limits, though its performance claims rest on the author's own benchmarks.
-
5
CNC Stone Carving - Carmelite Gothic carmelitegothic.com · Original Thinker88verdict
An insider's walk-through of the monks' CNC stone workflow, with real shop anecdotes and a neat pointing-machine analogy.
-
6
Official Claude Code Extension Compromise: Security Advisory | Phorion phorion.io · Original Thinker87verdict
A hands-on teardown of the backdoored Cursor Marketplace Claude Code extension, giving specific IOCs, the bugs that blunt the payload on each platform, and practical detection advice.
-
7
CircuitHub - Unveiling The Grid: PCBs made in America, twice as fast as Shenzhen circuithub.com · Original Thinker86verdict
CircuitHub's own unveiling of The Grid gives specific architecture and pricing details, though most performance claims come from the vendor itself.
-
8
Git for the US Code | Lexifina lexifina.com · Original Thinker84verdict
A product write-up with original validation data (361 of 365 changes reproduced) and a concrete retroactivity example, though the cited LLM studies are unsourced.
-
9
Google Search Is Less Than 50% of Alphabet's Revenue for the First Time Ever | Company Atlas kadoa.com · Original Thinker83verdict
A tidy, internally consistent numbers piece showing Search dipped to 49.7% of Alphabet revenue as Cloud doubled, with some historical figures unsourced.
-
10
Anthropic hacked to deliver remote access trojan to Claude developer laptops worldwide doublepulsar.com · Original Thinker82verdict
A detailed incident write-up with useful IOCs and remediation notes, though key figures like 5000 affected systems are unsourced.
-
11
H.264 is magic: a technical walkthrough of a remarkable technology. sidbala.com · Original Thinker81verdict
A clear, well-illustrated explainer of textbook H.264 concepts, made vivid by one real-numbers worked example.
-
12
OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why? minimaxir.com · Original Thinker81verdict
A data-backed case that open-weight competition, not subsidy, drives LLM price cuts, though some figures go unsourced.
-
13
AI Isn't Replacing SREs. It's Deskilling Them. newsletter.signoz.io · Original Thinker80verdict
A competent transfer of familiar automation-irony arguments to SRE work, with thin first-hand evidence and one unverified PagerDuty claim.
-
14
CSS-Tricks has a new owner! | CSS-Tricks css-tricks.com · Fresh Voice79verdict
The new owner's own announcement of buying CSS-Tricks, with a specific plan for ads, memberships, a course and slower publishing.
-
15
Northflank - Status status.northflank.com · Fresh Voice79verdict
Northflank's own incident notice urges users to rotate all secrets after unauthorised use of AI API keys, and gives a timeline and the remediation steps it has taken.
-
16
AI and the Fall? of the Creative Class apricitas.io · Fresh Voice78verdict
A data-driven look at shrinking US creative employment that argues persuasively for an AI role while acknowledging uncertainty and macro smallness.
-
17
Roblox downloaded 145 GB of models from our sites (including adult content) without asking open3dlab.com · Fresh Voice78verdict
A first-hand, log-backed account of a Roblox-labelled crawler scraping 145.5 GB of 3D models, which it then uses to argue that copyright is enforced unevenly.
-
18
Steve Blank Elon Musk Just Gave Europe a Reason to Replace American Technology steveblank.com · Fresh Voice72verdict
This is a provocative opinion piece tying Starshield restrictions and AI to European tech decoupling, but its key facts and revenue estimates are unsourced.
-
19
Home imaginationlibrary.com · Fresh Voice70verdict
A brief, accurate-for-today overview of the Imagination Library's book-gifting model, adding nothing beyond the program's own messaging.
-
20
Oregon Police Are Charging a Local News Outlet $2.5M for Records About One Cop theintercept.com · Fresh Voice70verdict
A pointed op-ed built on a fresh local records fight, padded with familiar federal FOIA anecdotes and stock reform proposals.
-
21
Genie 4: A new generation of world simulation deepmind.google · Fresh Voice69verdict
The primary announcement of Genie 4: native audio, better human and physics simulation, Gemini-guided planning, limited research preview.
-
22
bliki: LLM Voice martinfowler.com · Fresh Voice66verdict
A thoughtful practitioner's case for writing your own prose, with familiar arguments made concrete by personal practice and specific exceptions.
-
23
Python's experimental JIT to be removed lwn.net · Fresh Voice65verdict
A detailed news report on how PEP 836's rejection-by-feedback and withdrawal leads to CPython's experimental JIT being removed from main.
-
24
Mistral Large 4 (Le Chonk) is now open weight | Mistral mistral.ai · Fresh Voice63verdict
Mistral's own release notes give concrete benchmark, safety, license and pricing details, but nearly all performance claims are self-reported and unverified.
-
25
Apple's git still trusts the compromised eMudhra certificates emudhra.we-have-a-problem.com · Fresh Voice63verdict
A focused, demonstrated report that macOS's bundled git, curl and python3 still trusted compromised eMudhra certificates after 27.2, with a clear demand that Apple ship a fix.
-
26
Reps. Casar, Waters Lead Democratic Resolution Opposing A.I. Bubble Bailout for Big Tech, Commit to Protecting American People Instead progressives.house.gov · Fresh Voice63verdict
A press-release-style announcement of a Democratic House resolution that opposes AI bailouts and lists its provisions, framed with standard AI-bubble and 2008 rhetoric.
-
27
Anthropic illegally trained Claude with Stanford professors’ books stanforddaily.com · Fresh Voice63verdict
Student reporting that adds Stanford authors' voices and their opt-out lawsuit to familiar Bartz v. Anthropic background.
-
28
Blocks — Die Schweiz 1:1 in Minecraft blocks.manifold3d.ch · Fresh Voice63verdict
The original release page for a full 1:1 Minecraft Switzerland, with real technical detail and one odd residency restriction.
-
29
Survival LLM: an AI that still works when the internet doesn't brangus.io · Fresh Voice61verdict
A concrete, developer-written introduction to an offline survival AI with useful own measurements, though its accuracy test is small and unverified.
-
30
Meta Ignores Brazilian Court Order to End Bolsonaro Favoritism panamericandispatch.org · Fresh Voice61verdict
A competent wire-style report detailing activists' WhatsApp, Instagram and Meta AI search tests submitted to Brazil's electoral court, relaying their claims without independent verification.
-
31
I refuse to accept a censored Internet. Join Me at FOSDEM 27 talkweb.eu · Remix57verdict
A focused call for papers that argues credibly for treating circumvention, anonymity and messaging tools as one chain with shared threat models.
-
32
A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis lynchlab.net · Remix47verdict
An original benchmark by a dermatologist finds a 20-minute fine-tuned Inception v3 beat two frontier LLMs (62% vs 42%), though it still sent 18 of 30 melanomas home.
-
33
EMACS turns 50 technologyreview.com · Remix41verdict
A competent anniversary history of Emacs and its free-software legacy, built on well-documented accounts with few new details.
-
34
Why Is This Fingerprinting Script Creating Bioweapons? nullpt.rs · Remix40verdict
Original crawl-based research showing an anti-bot script that plants safety-triggering text to stop LLMs deobfuscating it, though its actual data collection is routine.
-
35
Websites Should Be Printable brennenputh.me · Echo36verdict
A friendly, light personal note on adding basic print CSS to a blog, built on standard techniques with a bit of first-hand experience.
-
36
Build Your Own Reasoning Model: A Full Video Course magazine.sebastianraschka.com · Echo32verdict
A promotional course outline listing which reasoning-model techniques are coded on Qwen3 0.6B, with a few specific design notes.
-
37
No Signs of AI Payoff Outside the Tech Sector | The Daily Spark apollo.com · Echo30verdict
Mostly a rewording of Apollo economist Torsten Slok's note, plus a short and familiar argument that AI lacks business judgment.
-
38
AI Promised Less Work. I Have Never Worked More. - Orestis Ioannou oioannou.com · Echo20verdict
Familiar AI-workload paradox made concrete by the author's own GitHub contribution statistics across roles and years.
-
39
Inside Wikipedia’s Cleanup Effort Against AI Text spectrum.ieee.org · Echo2verdict
With no extractable claims, this piece cannot be credited beyond a familiar framing of Wikipedia's fight against AI-generated articles.
-
40
IBM CEO says open weight models are now irresistibly good and cheap on-prem theregister.com · Echo0verdict
A headline feed with no assessable claims, adding no analysis or original content.
Scores are Expulp Verdict scores (0-100). Each link opens the full verdict with every claim weighed. Pieces are the week's Hacker News front-page articles from independent sites.