Achyuth Samudrala @achyuthnarayan

Physical AI | AI Infra London Joined July 2011

Tweets

8
Followers

40
Following

124
Likes

37

Nicolas Keller @Nicolas_Keller

3 days ago

Agents are coming for AI robotics research. We let an agent extract claims from 10 recent papers and turn them into experiments, including implementation & training. Then we gave it access to our farm, where it ran 6,500 real-world rollouts to test those claims. The infrastructure worked flawlessly. DM me if you or your agent want to run experiments in the real world or in sim, at scale :)

8 12 93 5K 52

View Details

Baseten @baseten

3 weeks ago

Inference engineering doesn't have a textbook yet. So we wrote one. 264 pages, free PDF and audiobook.

101 211 2K 6.0M 2K

View Details

Nicolas Keller @Nicolas_Keller

a week ago

Towards an agent-native CERN of AI robotics AI agents can now run the whole robotics research loop in hours: code, train, evaluate. But only in simulation. So we built a robot farm agents can control through an API. First test: can they replicate recent robotics papers? 1/17

9 14 70 4K 23

View Details

Nicolas Keller @Nicolas_Keller

2 weeks ago

π0.5 vs. MolmoAct2. Same task. Same object placements. Same training data regimen. Precise spatial placement lets us compare models side by side on any matching rollout in our Reality Check benchmark. Someone jokingly called it “Model Smash.” 👇 1/3

10 22 181 13K 133

View Details

Ai2 @allen_ai

3 weeks ago

MolmoAct 2, our fully open robotics model, takes the top spot for overall task success on the independent Reality Check benchmark after task-specific fine-tuning. Thanks @Nicolas_Keller for putting MolmoAct 2 to the test and sharing your methods and results with the community. 🚀

Nicolas Keller @Nicolas_Keller

3 weeks ago

The Era of Evals is coming to AI robotics. Today we’re launching Reality Check: our leaderboard & first public robot manipulation benchmark. 14,400 real-world rollouts. Four models. Multiple tasks & data regimes. Object placements. Confidence intervals. FR3 Duo stations. 1/10

60 71 687 119K 483

2 12 78 13K 17

View Details

David Rozenberszki @david_roz_

3 weeks ago

We're releasing PAW-10, a benchmark to give physical AI a reality check. Every object is scanned and paired as a real + sim asset, with placements in both guided to the millimeter. That's no accident. It's what you need to understand, break, and iterate on robot policies.

Nicolas Keller @Nicolas_Keller

3 weeks ago

60 71 687 119K 483

1 3 19 1K 5

View Details

Nicolas Keller @Nicolas_Keller

3 weeks ago

60 71 687 119K 483

View Details

Nicolas Keller @Nicolas_Keller

a month ago

Congrats to the @odysseyml team on Odyssey-3! The @pokeandwiggle team is so proud to have supported you in extending the model's capabilities to robotic manipulation. Excited to announce a close research collaboration with you guys! ❤️ 1/3

Odyssey @odysseyml

a month ago

Today we’re unveiling Odyssey-3, a big step forward for foundation world models. It can control robots, power humanoids, drive cars (on the roads of India!), train AIs, pilot drones, and even play video games. We can’t wait to see what intelligent systems it enables.