Reece Keller

Congrats to @KlemenKotar on this release. Physical world models are getting to the next level. You should really check out some of the amazing examples neuroailab.github.io/psi-website/bl… Fine-grained control for e.g. bottle-cap turning and needle-threading ... amazing!

Ok -- I've started a substack -- The Principle Investigator! My goal is communicating advanced takes on ideas in NeuroAI+. Here's the first substantive post: danyamins.substack.com/p/kolmogorov-c… Calculating the aesthetics of DNN models of the brain

Recently @hadivafaii invited me to discuss our work on intrinsic motivation at his awesome Sensorimotor AI journal club. Hadi wrote this excellent blogpost summarizing our discussion, check it out! sensorimotorai.github.io/2026/04/09/zeb…

If you're attending @CosyneMeeting, come check out our NeuroAgents workshop on Tuesday March 17. Speakers include: Omri Barak, Cristina Savin @lilwebian @rdkeller Caroline Haimerl @JonathanCKao @Ch0iHannah @xaqlab @srinituraga Yanan Sui @TrackingPlumes More details below 👇

Kids spend years playing with blocks, building spatial+arithmetic skills. Today, AI models just read. While AI research often conflates reasoning with language models, block-building lets us study how embodied reasoning might emerge from exploration and trial-and-error learning.

@xlr8harder @fleetingbits The prior encoded by a genome is in no way comparable to human-generated data. Sutton’s point was not that we should start from zero, but to challenge RLHF as a good prior for understanding the world since HF is not an objective nor well-defined reward function.

@fleetingbits I don't get the issue with pretraining on human-generated data. It's an effective way to bootstrap a prior on the world. Humans have a prior on the world too, via our genome. Why should we start from zero? Am I missing something?

Some thoughts on the Dwarkesh Richard Sutton interview: 1) Richard Sutton has internalized the bitter lesson to a very impressive degree. 2) He doesn't like pretraining because human set the data used in pretraining. He doesn't like post-training because humans set the curriculum. 3) He wants the agent to be able to be given a goal and then be able to loop to learn how to accomplish the goal on its own, just interacting with the world. 4) This involves the agent getting a progressively richer world model, related to its goals, which it is able to manipulate to accomplish its tasks. 5) I don't think that anyone at OpenAI, DeepMind or Anthropic would really disagree with this as the ultimate goal. 6) Whether, it is specialized models that interact in order to form an agent, with an interior training loop, or whether it's in context learning, or whatever. 8) I think the bigger issue with the interview was just that Dwarkesh wasn't familiar with Richard Sutton's way of thinking or talking. 9) Richard Sutton feels very connectionism, early AI, etc... and he understands the material, but he has a more focused worldview.

Here is our best thinking about how to make world models. I would apologize for it being a massive 40-page behemoth, but it's worth reading. arxiv.org/pdf/2509.09737

@RahulChandwaney Amazing work! Where is the top left photo from?

Inspiration VS Artwork

Over the past 18 months my lab has been developing a new approach to visual world modeling. There will be a magnum opus that ties it all together out in the next couple of weeks. But for now there are some individual application papers that have poked out.

Given the confusion around what RL does for reasoning in LLMs, @setlur_amrith & I wrote a new blog post on when RL simply sharpens the base model & when it discovers new reasoning strategies. Learn how to measure discovery + methods to enable it ⬇️ tinyurl.com/rlshadis

New paper: World models + Program synthesis by @topwasu 1. World modeling on-the-fly by synthesizing programs w/ 4000+ lines of code 2. Learns new environments from minutes of experience 3. Positive score on Montezuma's Revenge 4. Compositional generalization to new environments topwasu.github.io/poe-world [1/n]

@VictorTaelin Morphoceuticals

what are the most exciting developments taking place in biotech right now? companies or teams to watch for? what is the "this will solve everything" of today? CRISPR?

10/ Animal-like autonomy—flexibly adapting to new environments without supervision—is a key ingredient of general intelligence. Our work shows this hinges on 1) a predictive world model and 2) memory primitives that ground these predictions in ethologically relevant contexts.

9/ Finally, we show that the neural-glial circuit proposed in Mu et al. (2019) emerges from the latent dynamics of 3M-Progress agents. Thanks to my collaborators Alyn T. and @fel_p8, and to @xaqlab for his continued support! Paper link: arxiv.org/abs/2506.00138

1/ I'm excited to share recent results from my first collaboration with the amazing @aran_nayebi and @Leokoz8! We show how autonomous behavior and whole-brain dynamics emerge in embodied agents with intrinsic motivation driven by world models.