Luis

@DJiafei This is pretty cool, looking forward to the open code (and weights?)!

What if robots could think longer on harder problems without saying a single word?🤔 We introduce RD-VLA (Recurrent-Depth VLA): a latent, iterative reasoning architecture for robot control. ❌No Chain-of-Thought tokens. ❌No extra memory overhead. ✅Just reasoning—directly in latent space. 🧠🤖 Project page:rd-vla.github.io 👇🧵

Waypoint 1 launch rooftop launch party

meet alfred — our @NVIDIARobotics x @huggingface hackathon robot that helps makes an old fashioned by dropping the orange garnish in the glass - fine-tuned gr00t n1.5 on 50 teleop demos - trained and run on lerobot so-101 arms - generalizes to new drinks in the glass - bonus: running remote inference on @SpaceX's starlink since wifi was so bad (thanks @elonmusk!) code here: github.com/LF-Luis/HomeLe… @SeeedStudio #EmbodiedAIHackathon @drewborinstein @luifern0

If you’re a researcher at Meta/FAIR who worked on robotics/world models that’s affected by the recent layoffs, please contact me! We’re hiring!

We (Parallax Worlds) are VERY excited to host a robotics happy hour for SF @Techweek_ in our new space in South Park! Join us on Oct 9 with a curated group of robotics founders, engineers and investors in the heart of SF. Register here: partiful.com/e/6Sir9B3OIvSG…

@KarlPertsch That’s awesome! Will there also be a pi05_fast_droid_jointpos, or is pi05_droid trained on joint position action space?

Two big updates to openpi today: we are releasing pi0.5 and pytorch model code! This should be a straight model upgrade over pi0 in all aspects. See eg pi05_droid leading previous models in open RoboArena evals! :)