I want to make clear how crazy impressive this result is. We can now do bi-manual, dexterous manipulation across a wide range of tasks with barely any data on these skills coming from teleoperation. As we know, teleop does not scale! But turns out human video does! This means Optimus can now rapidly pick up new skills, with significantly less effort than before. What's next? You guessed it - RL :) Who would have guessed that humanoids are good for learning from humans 🤷‍♂️

I've seen the factories myself, the bots will be there! The AI team is pushing hard to make them useful :)

Dance mastery: 💯 Videos are one thing, live is a different beast altogether You can see how smooth and natural Optimus’ movements are We are extending this tech to quite a few more areas beyond this

Out of 16 teams, the winner of the @ucberkeley @CalHacks robotics track delivered MIRAI, an experimental human-robot interaction framework that enables natural motion / voice-based control of robots. Congrats to the top 3 teams, with students receiving prizes from @frodobots and @sigrobotics LeKiwi kits. Big thanks to the judges from @Tesla_Optimus @thebotcompany @Amazon FAIR, and more → @SOTA_kke @mihdalal @ZeYanjie @NiccoloFusai @micaelccarvalho

Had a great time judging the robotics track at Calhacks! Thanks for inviting me @BitRobotNetwork! Saw a lot of cool demos: from real steel on the G1 to robot deoderant sprayers, this session had it all Very exciting to see the next generation of robotics hackers combine AI with their own diverse backgrounds!

IROS Best Student Paper Award for Neural MP: A Generalist Neural Motion Planner! Congratulations to @mihdalal & @Jiahui_Yang6709 for leading this work at @SCSatCMU with @mendonca_rl, Youssef Khaky, @pathak2206. Check out our paper/code: mihdalal.github.io/neuralmot… Conventional motion planning solves each problem from scratch, demanding high computation and time. This work develops a data-driven approach that learns from large-scale simulated environments and expert trajectories to train a reactive generalist policy end-to-end, refined through lightweight optimization for safe deployment. Evaluated on 64 real-world tasks, the new method improves success rates by 23%, 17%, and 79% over state-of-the-art sampling, optimization, and learning-based planners.

woah another cool work by rohan! definitely worth checking this out

Proudest moment of my research career, Neural MP won Best Student Paper at IROS! We built a fast, reactive neural motion planner that works in the real world and ended up being used to be enable incredibly long-horizon manipulation tasks in ManipGen (mihdalal.github.io/manipgen/). This was a culmination over a year's worth of effort starting all the way in Spring 2023, with a super fun discussion with @pathak2206 :) Neural MP would not have been possible with out the incredible work put in by the entire team, especially @Jiahui_Yang6709 (who is applying for PhD Fall 2026 and would be a great candidate for any program!) Definitely check out our robot videos, code and paper here: mihdalal.github.io/neuralmot…

Congrats @Adithya_Murali_! This is awesome and very well deserved! Your work inspired a lot of mine during phd :)

Ya long tail gonna be rough for robots as well, probably worse, its gonna be a grind We have better ML tools now though

this might be the coolest looking version of the bot we have done so far

At this point we can just kind of “download” these skills into Optimus’s brain. Literally Neo from the Matrix 🥋 🤖

Going from phd to working at Tesla, this was one of the biggest changes. Don’t think I ever even thought about latency once in academia 😂

Scaling in sim feels like the right approach for the motion planning problem (no contacts to deal with). This is a great follow up work to my Neural MP paper with excellent results! Especially real-time reactivity looks very impressive!

Starting to look very futuristic at this point, but the next one is gonna be 🤯

All black is a vibe tbh This is the beginning

This is insanely cool, how does one get an invite @TheHumanoidHub