Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Hands Off: Detecting Model Weight Tampering
Stealthium introduces Rank-One Model Editing (ROME) attacks, where malicious Python code rewrites model weights in GPU memory to corrupt answers without touching disk, explains why CPU-focused EDR tooling misses them, and shows how its runtime sensor ties CUDA kernel launches and weight changes back to the Python code responsible.
13 min · 2,928 words
Cloud of suspicion lifted from NIST curves
John D. Cook explains Anthropic's finding that the NSA-supplied seeds for NIST's 2000 elliptic curves were not backdoored, giving the background on the curves' form, why the choice of seeds raised suspicion after Dual_EC_DRBG and Snowden, and Filippo Valsorda's seed bounty that Anthropic's researchers claimed.
2 min · 469 words
A benchmark map of the US economy
Mercor maps 29 high-quality agentic AI benchmarks onto the US Department of Labor's O*NET taxonomy and finds only 208 of 923 occupations have a task with at least five benchmark prompts, implying about 90% of American working hours go to tasks with little or no benchmark coverage. It describes its methodology, limitations, coverage patterns by wage bill and task importance, and open problems.
10 min · 2,373 words
Making Chicken Transparent Using Dorito DustModifying the optical properties of biological tissues with tartrazine
CasualPhysicsEnjoyer walks through a Nature paper on making live tissue optically transparent with absorbing molecules such as tartrazine (the yellow dye in Doritos), then recreates and explains the physics at home: refractive-index matching, complex permittivity, and why a yellow dye can open a window into chicken meat.
15 min · 3,512 words
Anthropic economists compare survey answers from US adults and Anthropic staff on how AI will reshape the economy. Both expect rapid progress, but staff expect much more: 91% versus 39% expect a Nobel-level AI discovery by 2030, and over half of staff versus 5% of US workers expect AI to do at least 80% of their own work by then.
19 min · 4,297 words
How employment is changing at Anthropic
Anthropic economists study how AI has restructured work inside Anthropic: hiring shifted away from junior technical roles as AI agent hours grew to outnumber employees six to one by August, output per person rose as staff adopted new tools, and human attention became the bottleneck. The paper ends with lessons for other employers.
23 min · 5,401 words
Working at the frontier shapes expectations for the future
Anthropic introduces two Oct 16 reports: how AI is changing hiring and work inside Anthropic (fewer junior hires, agents outworking employees, human attention as the bottleneck), and a survey showing staff expect far faster AI progress than US adults while reasoning about AI and jobs in a similar way.
2 min · 559 words
The math that OpenAI keeps for itself
Daniel Paleka on what OpenAI's flood of AI-proved math results conspicuously skips—especially cryptography—and why the gaps matter for research priorities before the field's incentives shift further.
5 min · 1,192 words
Basis researchers show how LLMs can invent a programming language, write an interpreter, and prove that the interpreter follows the language's rules—arguing that shipping a verified interpreter is becoming a practical unit of AI-generated software, not just generated code snippets.
19 min · 4,264 words
How worried should we be about extinction risk from AI?
Fields Medallist Timothy Gowers reflects on Jacob Tsimerman's pivot to AI safety and argues carefully about extinction risk, timelines, and what mathematicians can usefully do.
34 min · 7,782 words
Why are we even doing all this?A guest post on What’s new about the value of curiosity-driven work in an age of AI
Rikab Gambhir’s guest post on Terence Tao’s blog asks what research and learning are for when AI can already answer many technical questions—arguing for curiosity, taste, and the human practice of doing hard things even when tools get smarter.
6 min · 1,455 words
Bringing the magic: Google breaks open a bottleneck in quantum computing
Google Quantum AI reports in Science that magic state cultivation works in practice on its Willow processor, cutting errors about 40x versus no cultivation. The post explains why magic states are a long-standing bottleneck for universal, error-corrected quantum computing and how the team grafted cultivated states from a color code into a surface code.
3 min · 735 words
Neural Networks Are Already CleanA working hypothesis about neurons, generalization, and the ecology of circuits
Nick Cammarata of Goodfire argues neural networks are already 'clean': polysemanticity and messy circuits may be less fundamental than assumed, and studying neurons and circuit ecology points toward more interpretable structure than the superposition story alone suggests.
139 min · 32,079 words
Frontier Pre-training at Microsoft AI
Microsoft AI details pre-training MAI-Base-2, a 2.8-trillion-parameter MoE foundation model (80B active) trained on 16k GB200 GPUs over 38 days, which it says beats the open-weight models it tested on held-out evaluations at 4-12x better pre-training FLOPs efficiency, covering scaling ladders with source scaling, hyperparameter scaling laws, MoE and Muon optimizations, activation gradient clipping and native vision data.
12 min · 2,874 words
Brian Potter of Construction Physics looks at whether robotics can repeat AI's scaling playbook, explaining what robot training data is, how it is gathered through teleoperation, simulation, and other sources, and how large today's robot datasets are compared with those behind language models.
16 min · 3,630 words
Simons Foundation Launches $15M Effort To Help Mathematicians Navigate AI Era
The Simons Foundation, Anthropic and OpenAI commit $15 million in combined funding and resources to help the mathematics community adapt to AI. The foundation will run an open call for workshops and initiatives on evaluating AI-assisted results, learning new tools and setting shared standards, while the two labs provide AI tools and compute to supported researchers.
1 min · 314 words
CERN and Mistral join forces to accelerate AI for science
CERN and Mistral AI have signed a statement of intent for a long-term partnership on foundation models for science, starting with a pretrained Large Physics Model for collider physics trained on billions of simulated LHC proton-proton collisions to support tasks from particle reconstruction to analysis.
2 min · 553 words
Introducing Timestral: Mistral's time-series foundation models
Mistral releases Timestral, its first family of time-series foundation models in four sizes from 5M to 209M parameters. The largest ranks first on GIFT-Eval, TIME and fev-bench, delivers probabilistic multivariate forecasts with past and known-future covariates, runs efficiently on CPU, and ships under a modified MIT licence.
6 min · 1,385 words
Ultramarine 1 Flash: Scaling Non-Verifiable RL
Paper Instruments presents Ultramarine 1 Flash, an agentic model for professional knowledge work trained with large-scale reinforcement learning on GLM-5.3 Flash. It improves averages across six knowledge-work benchmarks while generating fewer tokens, and costs about $0.13 per task on their internal suite.
32 min · 7,459 words
The Public Pays to Produce Research. Shareholders Collect the Return.
SPARC releases its Cost of Knowledge accounting analysis and profitability report, which for the first time counts the public investment behind a journal article (typically over 95% of the total before submission) and tracks the billions paid to shareholders by the four largest academic publishers.
4 min · 877 words