
I gave a talk at ICLR 2026 about how we are scaling RL on frontier LLMs with 1T+ parameters, on experimental data from our physical lab at Periodic! Here's a rough recording of the talk:

Reinforcement Learner

I gave a talk at ICLR 2026 about how we are scaling RL on frontier LLMs with 1T+ parameters, on experimental data from our physical lab at Periodic! Here's a rough recording of the talk:

Introducing NovaAtom-Lite, an all-atom structure prediction model for drug discovery. It predicts the structures of proteins, nucleic acids, and the molecules they bind, including how they fit together. NovaAtom-Lite outperforms both IsoDDE and AlphaFold 3 on FoldBench antibody–antigen structure prediction (DockQ ≥ 0.23): NovaAtom-Lite: 79.07% (n=172) IsoDDE: 75.58% AlphaFold 3: 47.90% (n=167) Our evaluation uses 5 seeds × 5 samples, with the top-1 prediction selected by the model's own ranking score. Better structure prediction. A stronger starting point for your next experiment. Live today via Geodesic API and NovaDDE. 🔗 API: platform.geodesiclab.com/ 🔗 NovaDDE: agent.geodesiclab.com 🔗 Model Card: geodesiclab.com/news/introdu…

Beam’s model technical report is now available. Read the full report to learn about the methodology our team used to train Beam end-to-end in-house, here: reflection.ai/documents/beam…

Figuring out the mid-training recipe for this model was an incredible learning experience. There is lot's we can't share, but some little insights did make it into the blog post. One of my favorite ones is the usage of token efficiency plots when doing dataset ablations.
For the first time, we also added a mid-training stage before RL. We trained on several trillion tokens of code, scientific articles, blogs, and more, saving the scarcest, highest-quality data for last.

1/n Does scaling test-time compute on LLM self-verification keep paying off? 🤔 Not really. Our work shows scaling compute makes judgements more consistent, but not necessarily more correct 🧵 Paper: arxiv.org/abs/2611.00350 Blog: anirudhb11.github.io/blog/sc… Code: github.com/anirudhb11/scorpi…

Periodic Labs is heading to #SGLangSummit 2026! We’re building AI scientists and autonomous labs that learn from real-world experiments. This requires powerful models, robust infrastructure, and a tight feedback loop between ideas and experimental results. I’ll be talking about how we’re building RL infrastructure for physical experiments—from developing custom trainers and sandboxes from scratch to adapting SGLang to optimize RL workloads. Excited to see you all there! @sgl_project @radixark

Hi! this is Rohan Anil, I develop training algorithms among other things - my work trains various frontier models. At CoreAuto, I was fascinated by the problem of training on compute islands, eventually in space training. I hope you all enjoy this write up!🪐🛰️ coreauto.com/hitchhikers-gui…

New DeepMind Institute essay: "What is work for? Human wellbeing and the value of work in the AGI era." The empirical evidence shows that many people value their jobs beyond earning a salary, and the importance of agency, status, and meaning. At the same time, we investigate alternate sources of wellbeing, and the possibility that societal norms around work may change over time. with @JulianDJacobs @locross Read it here: bit.ly/wellbeing-work-agi

it's great we have more than 1 frontier lab, even more would be better
Excited to announce Gemini 4 Carbon, our next frontier model which is a step up from Argon across all capabilities! The benchmarks don't tell the full story -- we have been using it internally and it's been an upgrade for us.

Naive q: Why don't we see more LLM model releases with 10M context? Is it because they have to switch to non-quadratic attention?
Fern’s architecture, training, and serving support up to 10M tokens of context. Next-token prediction keeps improving as context grows, all the way to 10M tokens.

To channel my inner Schmidhuber, we found this softmax "bug" ~1.5 years ago and published in this paper arxiv.org/pdf/2603.11682 In reality, it's pretty hard to come up with a training scenario where this matters because gradient w.r.t. highly likely tokens is tiny anyway. Nonetheless I recommend fixing it -- one less thing to worry about! Check out @jon_barron 's work for an elegant fix!
Hi all, I have a new result I'm excited to share. It turns out that *all existing implementations* of softmax and cross-entropy (that's JAX, PyTorch, textbooks, etc) have a subtle bug --- two bugs, actually. A simple rewrite fixes both, and eliminates "slingshots" when grokking:

Not sure if there's any recording for my CoLM talk but you can find the slides here. agarwl.github.io/slides/cont… Posting couple of slides which look directionally good (to me)
Giving a talk at CoLM today in 30 mins, sharing some of our work at Periodic and and where it's heading.

Simple rewards, emergent complexity: give the agent something simple to optimize in an environment with a lot to learn about
David Silver is one of the all-time greats. He was the creator of AlphaGo, which changed the course of history by showing that neural nets could achieve superhuman performance for the first time. He also has a unique intuition and knack for forecasting when things will happen in AI. With computer Go, he predicted the Lee Sodol outcome 10 years before, right to the year. He got the year right for Starcraft and won a bet inside Deepmind. Same for AlphaProof. Same for the IMO. David's prediction for superintelligence: 3 years. !!! He has a strong sense that now is the time to build a superleaner, and that superlearners are now possible, thanks to the enormous progress in compute, simulators, autoresearch, and the sophistication of technology in the stack since the AlphaGo days. What is a “super learner”? A super learner discovers all knowledge from its own experience. Start from nothing (no internet pre-training), and go beyond human knowledge. Drop it into any stream of experience, simulated or physical, and the superlearner learns. He expects the progress will be rapid. When you get the recipe for a superlearner just right and remove humans from the development loop altogether, a superlearner can learn extremely quickly. Just look at AlphaGo. AlphaGo took two years to develop with humans in the loop. When David and his team at @GoogleDeepMind got the formula for AlphaZero correctly, it surpassed the human-developed algorithms in just a few hours. Human data is a shortcut, but you have to let go of the human data to shoot beyond it. In January, David left GDM to found @IneffableLabs, a research-led lab with one mission: build a "super learner" that discovers all knowledge from its own experience, and make first contact with superintelligence. He joins us with co-founder Lasse Espeholt, who built IMPALA and SEED RL, the foundations of scaled-up RL, and now runs Ineffable's compute. A few things we cover in this conversation. — What a super learner is: depth (start from nothing, go beyond human knowledge) and breadth (drop it into any stream of experience, simulated or physical). The human brain is the only existence proof — Why they're giving up human data: AlphaGo took two years with humans in the loop. AlphaZero outperformed it in a few hours. "You have to let go of the human data to shoot past the competition" — Crucibles: simulated worlds with enough emergent complexity (physics engines, millions of interacting agents) for an agent to rediscover technology, science, maths and language on its own — Simple rewards, emergent complexity: evolution optimized fitness and got all of life. Don't breadcrumb the objective, give the agent something simple to optimize in a world with a lot to learn about — Go as the existence proof: one agent is the most boring game imaginable, two agents is "an infinite onion of layers of complexity" — Why human language is probably not the endpoint for reasoning, and how training on it bakes in a flat earth and the four humors — Can superintelligence be built in simulation? If the sim is Turing complete, computation, economics, language and maths are all expressible, and it happens 10-100x faster and safer — Alignment: language-based monitoring is reaching its limits as agents learn to say one thing and do another. Shadow actions, superintelligences monitoring superintelligences, and building a real science of how they interact peacefully — Accelerate in simulation, pace in the real world: Ineffable will gate real-world deployment on evidence that it goes well — Timing as the determinant: Silver called Computer Go to the year, then StarCraft, then the IMO. He thinks this is the moment — Predictions: True superintelligence within three years. Timestamps: 0:00 – Introduction 1:09 – What held true from Go to today, and what the LLM era changed 3:35 – Why leave DeepMind to start a new lab 5:58 – Why "Ineffable" (and the secret second reason) 7:03 – Lasse's path: scaling RL at DeepMind and Google Brain 9:31 – Defining a super learner and first contact 12:35 – Super learners vs. LLMs that make discoveries 13:59 – Day to day: the RL stack, crucibles and human-journey milestones 16:13 – Don't breadcrumb the objective: simple rewards, emergent complexity 21:16 – What a crucible looks like 24:14 – Go: one agent is boring, two is infinite 25:41 – Will agents rediscover language? 27:22 – Removing human priors: AlphaGo to AlphaZero 32:05 – Keeping human bias out of the crucibles 33:35 – Confidence, timing and super forecasting 36:24 – Is compute the driver? 38:13 – The ceiling on simulation and Rich Sutton's objection 43:44 – Alignment and the science of superintelligence 47:45 – Monitoring agents that don't think in English 51:01 – Pacing the frontier 52:54 – The compute setup: Vera Rubin on GCP 55:18 – Training, inference, simulation and RL scaling laws 57:40 – IMPALA and SEED RL 1:00:58 – 20 people and the compute-per-person ratio 1:04:49 – A research-led culture 1:05:47 – Predictions

I hope other American companies follow suit
1/ muse spark open is the most capable US open-weight model to date. 816B total params, 42B active. 1M-token context. natively multimodal. built for coding and agentic work like long-horizon tasks, hard software engineering, multi-step reasoning. america! 🇺🇲

More often than not, whenever my hero RL run is behaving weirdly, it's an infra issue 🫠

Watched Digger, phenomenal. Can't believe it bombed while cliched Spiderman-part-1000 the biggest movie this year. Welcome to Level 4! @TomCruise gave the performance of his life.