Kia

The day a blind man sees. The first thing he throws away is the stick that has helped him all his life

Huge news: @prior_labs has signed a definitive agreement to be acquired by @SAP. €1B+ invested over four years to build a globally-leading frontier AI lab for structured data — in Europe, in the open. Independent entity. Same team, same mission, same open models. A massive boost to what we can do. The mission just got accelerated. Founders’ statement: priorlabs.ai/blog-posts/pri… (Deal subject to regulatory approval; terms not disclosed.)

enjoy that sunset while you can—soon a swarm of superintelligent AI agents will be able to appreciate it more rapidly and efficiently than you ever could

Our book on Generative AI and Stochastic Thermodynamics can be pre-ordered with a 20% discount until July 31 2027. (All proceeds from the authors will be donated to the African Institute for Mathematical Sciences).

New paper: "Agentic Forecasting using Sequential Bayesian Updating of Linguistic Beliefs". Our system (BLF) matches human superforecasters on ForecastBench, and beats all the top methods (GPT-5, Cassi, Grok 4.20, and Foresight-32B). 🧵

@rsalakhu @subail @chuckjhoover @asenkut @FHaskaraman Congrats! BTW you might find my recent paper of interest... arxiv.org/abs/2604.18576

Modern deep networks are often trained at the #EdgeOfStability, a regime where dynamics are locally unstable, nearing chaos. Yet generalization improves, defying the wisdom of classical optimization. We now theoretically explain this central puzzle: arxiv.org/abs/2604.19740. 👇

1/ New work (with @hayou_soufiane and Eric) on Expert Specialization in MoEs: arxiv.org/abs/2604.09780 We show that specialization arises from the similarity of hidden states, which is *complicated* e.g., optimizer dependent (adamw v.s. muon) and not always understandable 🙄

We released "The Newton--Muon Optimizer" . We show that Muon is secretly an implicit Newton method, and use this insight to build a better one. 1/n Paper: arxiv.org/abs/2604.01472

The Illusion of Learning from Observational Data: An Empirical Bayes Perspective Bohan Wu, Sebastian Salazar, Donald P. Green, David M. Blei arxiv.org/abs/2604.08853

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing

The NeurIPS 2026 submission site opening has been moved to April 15th. Note that all tracks (Main Submission, Evaluations & Datasets, and Position Papers) share the same dates. The final submission deadlines remain unchanged: • Paper Abstracts: May 4th (AOE) • Full Papers: May 6th (AOE)

Make no mistake: destroying world-class universities, like the US just did with Sharif University of Technology in Tehran, isn't just an attack on Iran but it's literally an attack on all of us, on all of humanity. It's not Iran that "won" when Maryam Mirzakhani made her discoveries that won her a Fields Medal: it's all of mathematics, and everything mathematics is used for. Human progress won, technology won, we all won. It's the same type of stuff the Mongols did during the sack of Baghdad and their destruction of the House of Wisdom: we ALL lost something irreplaceable back then, entire fields of human knowledge set back. That's what bombing a university does. It doesn't just destroy buildings. It destroys us, all of us.

Sharif University of Technology (founded as Arya Mehr University in 1966) is an icon of modernization and progress in Iran. Its alumni include the first woman to win the Field Medal in Mathematics, Maryam Mirzakhani. It has been a national symbol of achievement, gaining international recognition for the quality of its graduates, large number who have been admitted into the very best engineering programs in the West. The aim of this kind of wonton destruction could only be the nation of Iran itself.

SAEs fail at OOD tasks. Why? Features in superposition are linearly representable but not linearly accessible. Instead of discarding sparse coding, we embrace the geometry of superposition and use methods equipped to handle the nonlinearity it induces.

You may think discrete distillation is fundamentally flawed, you are (surprisingly) wrong. 🤯 Meet Discrete Moment Distillation (D-MMD). It is a new method that brings fast, few-step sampling to discrete diffusion models! 🧵👇

never in my lifetime i would have thought this image would exist

incredibly bullish on the future of tech + AI in London. just to name a few: • OpenAI just announced (last week) that London will become its largest research hub outside San Francisco • Anthropic kicked off a 100+ person hiring spree across London and Dublin in 2025 • xAI set up shop in London in early 2025, based in X's former Piccadilly office, led by former DeepMind researcher Toby Pohlen • Microsoft hired 24+ researchers from Google DeepMind for its London AI hub • Google DeepMind announced its first automated research lab in the UK (opening 2026), focused on discovering new materials using AI and robotics • Perplexity committed £80M to expand London offices • Groq is opening its first UK data centre in London • Cursor chose London as its European HQ

🚨🚨 New paper on flow-matching value functions Last year, we showed training RL value functions with a flow-matching loss achieved SOTA results. But why does it work? And what could it possibly tell us about other things that have nothing to do with VFs or even RL? Short answer: iterative compute used correctly can address feature plasticity in continual learning! 🧵⬇️

How are mathematicians facing the wave of rapidly advancing AI-for-math capabilities? Jeremy Avigad (CMU prof and co-author on the original 2015 system description paper for Lean) just posted a paper with his thoughts in the wake of the Math, Inc. announcement on sphere packing. andrew.cmu.edu/user/avigad/Pa… There are a lot of interesting passages in here, including a bit of the back story of the Math, Inc. bomb drop and how it was initially received by the humans working on the formalization project. But, as for how mathematics proceeds, here's the key last passage: "We need to remember our strengths: mathematicians are problem solvers and theory builders extraordinaire. Rather than fight the use of AI in mathematics, we should own it. It is not enough to keep up with current events and design benchmarks for AI researchers; we need to play an active role in deploying the technology and molding it to our purposes. We also need to learn how to raise our students with the wisdom to use the new technologies appropriately, and we need to be careful that we still manage to impart core mathematical intuitions and understanding. Figuring out how to use AI effectively to achieve our mathematical goals won’t be easy, but mathematicians have always embraced challenges—indeed, the harder, the better. If we face AI head-on and stay true to our values, mathematics will thrive. We just need to show up and get to work." The next few years should be a golden era for mathematics. For those of us working on the frontier, I hope we do well by our mathematician colleagues.