Co-Founder & CEO @AntimLabs // prev: @columbia
SF
Joined September 2015
- Tweets389
- Following487
- Followers2.2K
- Likes1.9K
Shrey Kothari retweeted
it's time for more interesting robotic evals:
hooked astra and fable 5.1 to a franka arm each and let them play connect 4 in mujoco.
the basic camera observations in -> EE poses + gripper commands out -> cartesian IK -> sim joint control loop. actual contact and gravity for every pickup and drop.
started both at high reasoning; resumed at medium after a per-player budget cutoff. fable won with a diagonal on move 25. $17.40 in API calls.
thought about chess, but this felt like a better first try. chess next!
we’re getting awfully close
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
Shrey Kothari retweeted
Some highlights:
• Since DeepMind’s foundation in 2010, the constrained yet rich worlds of games have played a critical role in understanding intelligence
• the spirit of exploration that succeeded in games had profound impacts for other AI systems: AlphaFold applied these foundations to help solve the 50-year grand challenge of protein structure prediction
• For AI research, this is a golden opportunity... demands precisely the capabilities we believe are essential for frontier AI:
• Continual learning
• Memory
• Long horizon planning
• Complex multi-agent dynamics
• Games have always been a mirror for intelligence
@demishassabis @amoufarek @GoogleDeepMind get it.
Games have been an important testbed for our AI research for over 15 years. 🎮
From mastering Atari to reaching Grandmaster in StarCraft II, they have driven some of our biggest AI breakthroughs.
Our work with SIMA taught agents how to understand 3D worlds, but learning to navigate real human dynamics takes a living, persistent universe.
Through our research partnership with @FenrisCreations, we’re exploring how to tackle open challenges in AI:
🔵 Continual learning to acquire new skills without forgetting past knowledge.
🔵 Deep memory systems that store and retrieve information far beyond today’s context windows.
🔵 Long-horizon planning over weeks, months, or years.
🔵 Multi-agent dynamics spanning cooperation, negotiation, economics, and emergent behaviors.
Our long-term goal is to use AI to discover entirely new gameplay experiences in partnership with game developers – making games more accessible and personalized – while applying what we've learned to problems in the real world and scientific discovery.
Find out more → goo.gle/4xcwzOH
ALT A graphic featuring the Google DeepMind logo and the text "From Atari to EVE Online: Building on 15 years of game AI research." To the right of the text, a grid illustrates the company's progress through four major gaming milestones: Atari, Go, StarCraft II, and EVE Online.
We’ve been building infra for robotics and defense autonomy companies spanning across indoor/outdoor/city-level and water sims and would love to get into the weeds of sims and scaling them to close the robot data gap and s2r. Please reach out!
we're working on sim infra for physical ai at @AntimLabs and want to meet more people thinking deeply about sims
especially interested in sim2real, contact/dynamics modeling, system ID, DR, r2s2r, scalable scene generation, synthetic data, and sim evals etc.
researchers / PhDs please reach out. would love to compare notes.
Shrey Kothari retweeted
That’s a wrap on ACTUATE 2026 by @foxglove
Great couple of days sharing what we’re building at @AntimLabs - the world layer for physical AI and meeting incredible folks pushing the frontier in robotics and embodied intelligence.
killer team, insanely ambitious mission
Today, we're introducing @discoveredmat . We build AI scientists that discover new materials for semiconductor chips.
We’re starting by releasing our work - hundreds of new materials discovered using frontier AI models, along with our benchmark for tracking progress in this domain, Material Discovery Bench.
To pursue our mission, we have raised $9 million from @LightspeedIndia (lead investor) , @ycombinator , @peakxvpartners , @paulg , @gokulr , @trq212 and many others.
🧵
alright who’s building ikea for robots
A robot dog you’re actually supposed to take apart.
China’s MirrorMe built Black Panther X as a modular, open quadruped platform rather than another closed robot dog.
200+ components, 12 joints, open interfaces + code, 4 m/s speed, 33 Nm joint torque, and support for third-party sensors and payloads.
You can assemble it, modify it, program it, and build on top of it.
Shrey Kothari retweeted
Simulation asset generation is going to massive for scaling simulations.
Codex / Claude Code are not trained on this kind of 3D sim asset code gen. @AntimLabs is catalyzing building what used to take days in minutes.
Faster asset Gen -> Faster sim creation -> more roll outs / more sim data -> better policies -> deployed robots!