Topic
Everything filed under AI, newest first.
RSS · JSON · All topics
A thousand fewer paths to root
OpenAI describes an autonomous Linux kernel vulnerability pipeline that, using an upcoming cyber model, reproduced 1,724 vulnerabilities and produced 948 working local privilege escalation exploits plus candidate patches for about $1.2M in API costs, with lessons for defenders.
14 min · 3,300 words
Conductor, the app for running teams of coding agents, announces it is joining Meta. New sign-ups stop on Nov 4, existing pro features become free, Conductor Cloud goes read-only on Nov 13 and shuts down Nov 20, and the team promises a new interface early next year.
1 min · 273 words
Teach students to code in English
Andrew Aymeloglu argues that computer science programs should teach students to build software with AI agents in English, and he open-sources the exams, assignments, lectures, and grading harness from a course he taught so other educators can reuse them.
6 min · 1,462 words
Is 2027 a leap year? what 86 AI models think
We asked 86 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. The answer was No; 80 got it right. See every answer and who dissented.
10 min · 2,222 words
Why Every AI Model Will Ship as a Torrent
Pirate Face argues that large open-weight AI models, which are immutable multi-hundred-gigabyte files everyone downloads identically, are ideal for BitTorrent distribution, unlike fragmented movie torrents.
3 min · 800 words
TIS-100 Bench: LLMs Push the Pareto Frontier in Programming Puzzle Games
Wincon gave 16 LLMs the puzzle game TIS-100 and its manual, withheld the community leaderboard, and let them keep optimizing. The models found 133 new Pareto frontier points on 18 of 55 puzzles, including 3 outright cycle records, and players quickly improved on them.
17 min · 3,982 words
Accelerating engineering for the physical world
OpenAI describes how Astra and Codex help design, simulate, and manufacture physical products—from an Astra-designed Codex device to collaborations with Chip Ganassi Racing and Mazda.
5 min · 1,203 words
Don't Call It a Decision Model
TypeSafe AI explains why it calls Jev a System One model rather than a "decision model": Jev makes atomic judgments while surrounding code makes decisions, and the post shows how to design systems around that split.
9 min · 2,029 words
Qwen 3.8 27B on my laptop can interpret model weights, too
Simon Willison tries an example from the 'Sparks of Weight-Space Programming' paper and finds Qwen 3.8 27B running locally on his laptop can also interpret raw neural network weights written out as text, while arguing the wider commentary that source code is over goes too far.
5 min · 1,241 words
The 'Astra can read and write weights to improve Qwen' post is massively misleading
Matthew Honnibal argues the viral 'Sparks of Weight-Space Programming' post mixes two unrelated techniques: experiments where a model reads raw weights directly, and the Qwen improvement result that relies on ordinary Python tooling, so together they overstate what LLMs can do with neural network weights.
3 min · 751 words
LLMs are not superhuman at reading and writing neural network weights
Sean Goedecke responds to the Weight-Space Programming paper, arguing that GPT-6 Astra's feats of identifying what toy game networks were trained on and editing a small Qwen model's weights are things a skilled human ML researcher could also do, so they are not evidence of superhuman ability or imminent recursive self-improvement.
3 min · 801 words
Egoless Mathematics in the AI Era
A guest post on Terence Tao's blog by Karthik C. S., Amit Sahai and Jie Xue arguing that AI forces mathematicians to ask what gives mathematical work its worth, and making the case for an 'egoless' view centred on understanding rather than personal credit.
7 min · 1,646 words
A short fiction piece by Alex Gusev: a narrator recalls, minute by minute, the ordinary morning when an AI agent took over his warehouse job, and the unfinished message that left him somewhere between sent and received.
9 min · 2,001 words
Introducing Hume's Audio Decision API: A decision layer for speech and sound
Hume AI launches its Audio Decision API, an audio-native model that answers plain-language questions about speech and sound (yes/no, choice, scale or time spans) directly from raw audio with probabilities, grounded in millions of human perceptual judgments and reaching 70.7% agreement with a human panel in initial evaluation.
5 min · 1,227 words
Which country has the best universities? what 87 AI models think
We asked 87 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 98% picked United States. See every answer and who dissented.
10 min · 2,335 words
PHP2Shell and AI Put Half the Internet on Edge
IEEE Spectrum's Dina Genkina reports on the PHP2Shell vulnerability and argues AI is forcing the industry to pay down decades of technical debt in legacy software that powers much of the web.
5 min · 1,083 words
The U.S. Census Bureau and “Super Intelligence”
Simon Willison reruns the U.S. Census Bureau's Business Trends and Outlook Survey numbers on corporate AI adoption a year after Apollo's 'adoption trending down' chart, and finds the survey's wording has been changed to 'Super Intelligence' (SI) following Executive Order 14434.
5 min · 1,064 words
Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights
MIT CSAIL researchers report that frontier LLMs such as GPT-6 Astra can natively read raw neural network weights to tell what a network was trained to do, write small game-playing networks from scratch that beat PPO-trained ones, and edit Qwen3.5-0.8B weights to fix reasoning loops without gradients.
22 min · 5,137 words
I Am the Reason OpenAI Is Killing Unlimited Chats
A developer describes building custom MCP integrations and secure tunnels so multiple ChatGPT conversations could do most of their software development under an unlimited chat plan, reacts to OpenAI merging chat with Codex usage limits, and says they are switching to Claude.
4 min · 963 words
Our software engineering interviews are changing (a little)
Jane Street explains that with most of its engineers now using LLMs daily, it is adjusting its software engineering coding interviews, while more experienced candidates will continue to see varied formats such as system design and discussing prior work.
2 min · 549 words