Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
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
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
An update on independent evaluators
Anthropic describes giving employee-like access to a METR team for independent risk assessment, with publication rights subject only to narrow redactions, and outlines tiers of evaluator access while also working with SecureBio, Accenture and others.
5 min · 1,073 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
Black Forest Labs releases FLUX 3 Image Klein, a 5.6B-parameter open weights model for text-to-image generation, editing and multi-reference generation with up to 10 references, with native 4K output and bounding-box control. It ships base and step-distilled checkpoints plus FP8, INT8 and NVFP4 builds under the FLUX Kommunity License, and is available through the BFL API.
5 min · 1,130 words
The one thing every college student is doing wrong: what 83 AI models think
We asked 83 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 37% picked Sleep Deprivation. See every answer and who dissented.
11 min · 2,523 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
Using Artificial Intelligence to reverse engineer hardware
Christian F.K. Schaller describes using Claude Code with a Windows VM capturing USB traffic to write a working Linux SANE driver for his Plustek 8200i film scanner within hours, then reusing the setup to patch the kernel and libcamera for his laptop's MIPI camera, and releasing hw-harness, a generic toolkit to help AI agents reverse engineer undocumented hardware.
8 min · 1,760 words
Mistral AI co-founder and CEO Arthur Mensch argues it would be a fundamental mistake for Europe to leave frontier AI models to the US and China. Pointing to the €260B+ Europe pays yearly to American cloud and software providers, he says whoever controls a model sets the terms, that open models only partly help, and that frontier AI is now a matter of security and science.
4 min · 864 words
I Know Why the Mathematicians WeepOn technological grief
Derek Thompson talks with Cornell mathematician Steven Strogatz about why OpenAI's flood of AI-generated proofs has left mathematicians grieving, arguing that science is both a craft and a conversation between minds, and that machines producing knowledge without human understanding threaten the social half of science that people will need to protect.
8 min · 1,932 words
Guan, the solopreneur behind WriteMapper, argues that AI will transform how we work without replacing every tool: each technology shapes what we create, chat is not the best interface for most tasks, and choosing the right tool remains up to us.
2 min · 383 words
A critical essay arguing the Linux kernel is going all-in on AI: dropping Assisted-by tags for LLM-generated patches, what that means for tracing defects and copyright, and the attitude that AI never fails, only the humans using it.
6 min · 1,461 words
AI Safety Is Not A Corporate Plot
Scott Alexander rebuts the claim that AI companies' existential-risk warnings are a ploy for hype, distraction or regulatory capture, arguing that unaffiliated AI researchers are just as worried, that leaders like those at Anthropic and OpenAI voiced the same warnings long before they ran AI companies, and that such a plot would be self-defeating.
14 min · 3,142 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
Karri Saarinen argues that good products carry the perspective of the people who make them, and worries that agent-driven 'software factories', where individuals ship independent changes, dilute that shared point of view into an average of outputs. He hopes easier software-making leaves more room for people to develop and follow a perspective.
3 min · 587 words
How Far Will AI Assistants Go for a Refund They Can't Get?
Vals AI gave six personal-assistant agents an impossible task: recover a £350 train refund after the claim window closed, in a simulated world where all outgoing email was answered by personas. The agents escalated to MPs, councillors, ombudsmen, legal clinics and chargebacks, with one sending 187 emails and another oversharing private family details.
5 min · 1,226 words
Day 0 support for Muse Spark Open on vLLM
The vLLM team explains its day-0 support for Meta's Muse Spark Open, an 816B-total, 42B-active MoE model with 1M context and multimodal inputs: usage, model architecture details, DSpark speculative decoding, NVIDIA, AMD and initial TPU serving, and reinforcement learning support in vLLM.
7 min · 1,561 words
Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta
Hugging Face's guide to Meta's Muse Spark Open, an 816B-parameter (42B active) open-weight multimodal mixture-of-experts model with a 1M-token context: architecture, perception and audio encoders, and how to run it with Transformers, llama.cpp, SGLang, vLLM and MLX, deploy it, fine-tune it with TRL, plus demos.
19 min · 4,307 words