The impact @karpathy has on the AI community is fascinating. He’s almost like an operating system: he surfaces important problems, directs collective attention, and makes the excitement contagious enough that people shift what they work on. All through sheer curiosity and range.

Our official main stage brings together 1,000+ of Canada's top builders from around the world. Featuring @tobi @andrewgordonmac @nickfrosst @lucyhargreaves4 @Sirupsen @MaxBrodeurUrbas @ElainaYallen @EliotPence @mitrymin @Alex_Danco and more. Limited in-person tickets. Request access ➡

Early on, we had so many hypotheses about the right memory structure for agents. Knowledge graph? Vector DB? Then we decided, screw it, just put it in a text file. For now. Messy, but kind of elegant. Fast forward to 2026: Openclaw lives in Markdown :)

We trained our models to solve problems with objective answers. But can we build models that solve problems where success is subjective, messy, and human? The latter is even more impactful imo. Agree with Percy: simulation is the next frontier for AI.

I think it’s pretty clear that simulation is the next frontier for AI. The most impressive feats of AI to date are when we have a clear environment + reward, whether it be beating Le Sedol at Go, winning an IMO gold medal, or writing entire apps from scratch. In these cases, the RL algorithm can try different actions, and observe the well-defined consequences in the safety of a docker container. But what about messy real-world situations involving people? The rewards are unclear, the stakes are high, and you can’t experiment in the real world. But these situations are precisely where the next big opportunity in AI is. To crack this, we need to *simulate* society (“put society into a docker container”). Concretely, this means building a model that can predict what will happen in any given situation (real or hypothetical). If we can do this, we are only limited by our imagination: predict the future, optimize for better outcomes, answer hypothetical (“what if”) questions. Ultimately, this goes beyond making better decisions, but it’s about giving us a better understanding of ourselves and the world. Simulation is the whole enchilada. And this is exactly the research that @simile_ai is working on. Read more here:

Our work on generative agents showed that it's possible to accurately simulate human behavior by capturing rich information about real people. @joon_s_pk spoke with @Nature about how our work builds on this research, and how enterprises can now use simulations to test decisions before they hit the real world:

Exciting to see @WSJ cover what we’re building at @simile_ai. It’s great to see the technology we developed in the lab making real world impact alongside foundational institutions like CVS and Gallup. Nothing like frontier research meeting real PMF!

Simile is increasing decision-making capacity and decreasing research time. We are proud to work with partners like @CVSHealth! 🤝 Projects that took months are now taking hours, and studies are closer resembling human behavior than prior self-reported research - all with the goal of providing the best products and services for customers at companies like CVS Health, where the customer is centered in every decision.   Read more in the CVS Health whitepaper below.

I still remember the excitement in 2023 when Stanford Smallville was launched. It was the largest multi-agent sim back then - yes, 25 bots felt like a lot. Today it's the "Bigville" moment. We are seeing a nascent, massive-scale alien civilization sim unfolding in real time: orders of magnitude more agents, way higher IQ, in-the-wild access to the internet, backed by the full arsenal of MCPs. What can possibly go wrong?

Announcing DreamDojo: our open-source, interactive world model that takes robot motor controls and generates the future in pixels. No engine, no meshes, no hand-authored dynamics. It's Simulation 2.0. Time for robotics to take the bitter lesson pill. Real-world robot learning is bottlenecked by time, wear, safety, and resets. If we want Physical AI to move at pretraining speed, we need a simulator that adapts to pretraining scale with as little human engineering as possible. Our key insights: (1) human egocentric videos are a scalable source of first-person physics; (2) latent actions make them "robot-readable" across different hardware; (3) real-time inference unlocks live teleop, policy eval, and test-time planning *inside* a dream. We pre-train on 44K hours of human videos: cheap, abundant, and collected with zero robot-in-the-loop. Humans have already explored the combinatorics: we grasp, pour, fold, assemble, fail, retry—across cluttered scenes, shifting viewpoints, changing light, and hour-long task chains—at a scale no robot fleet could match. The missing piece: these videos have no action labels. So we introduce latent actions: a unified representation inferred directly from videos that captures "what changed between world states" without knowing the underlying hardware. This lets us train on any first-person video as if it came with motor commands attached. As a result, DreamDojo generalizes zero-shot to objects and environments never seen in any robot training set, because humans saw them first. Next, we post-train onto each robot to fit its specific hardware. Think of it as separating "how the world looks and behaves" from "how this particular robot actuates." The base model follows the general physical rules, then "snaps onto" the robot's unique mechanics. It's kind of like loading a new character and scene assets into Unreal Engine, but done through gradient descent and generalizes far beyond the post-training dataset. A world simulator is only useful if it runs fast enough to close the loop. We train a real-time version of DreamDojo that runs at 10 FPS, stable for over a minute of continuous rollout. This unlocks exciting possibilities: - Live teleoperation *inside* a dream. Connect a VR controller, stream actions into DreamDojo, and teleop a virtual robot in real time. We demo this on Unitree G1 with a PICO headset and one RTX 5090. - Policy evaluation. You can benchmark a policy checkpoint in DreamDojo instead of the real world. The simulated success rates strongly correlate with real-world results - accurate enough to rank checkpoints without burning a single motor. - Model-based planning. Sample multiple action proposals → simulate them all in parallel → pick the best future. Gains +17% real-world success out of the box on a fruit packing task. We open-source everything!! Weights, code, post-training dataset, eval set, and whitepaper with tons of details to reproduce. DreamDojo is based on NVIDIA Cosmos, which is open-weight too. 2026 is the year of World Models for physical AI. We want you to build with us. Happy scaling! Links in thread:

TMI from the original Smallville sims. Maria and Klaus had a crush on each other. In half the simulations, Maria asked Klaus out; in the other half, Klaus did. Klaus == disaster. He invited her to an academic reading group for Valentine’s. As a fellow geek, I prayed for Maria.

Happy Valentine’s Day from Team Simile! Fun fact: In our founders’ 2023 paper, @joon_s_pk and team introduced the concept of generative agents to the world by simulating a town of 25 agents… one of who was planning a Valentine’s Day party. The agents autonomously spread invitations to the party over the next two days, made new acquaintances, and asked each other out on dates to the party. This town of agents, Smallville, along with all the flirty festivities, was instrumental in creating the field of AI-based simulation. Love is in the air! 🎈