Robots are getting more powerful hardware all the time. Their software is arguably more important as it can enable them to take on tasks they haven’t encountered before and evolve. Take ASENA for instance: it is a coding agent that writes and fixes the programs as it goes. It self evolves and saves what works as code for the next task, so no retraining is needed.
Our G1 robot found its way to the pantry, picked up a Pringles can, and handed it over. No pre-built map. No training for this task.
Behind it is a coding agent, writing, running, and fixing the robot’s programs as it goes. And it self-evolves: what works is saved as code for the next task, no retraining needed. This is ASENA.
Joint work with @Isabella__Liu, @edmund_bu, Johan Bjorck, @yin_hongxu, @zhengyiluo, @jankautz, @DrJimFan, @yukez, and @Sifei30488L. 🧵👇
— An-Chieh Cheng (@anjjei) October 14, 2026
As the researchers explain, ASENA gets better by commits. A G1 robot using this approach found its way to the pantry, picks up a can and hands it over with no pre-built map. The agent relies on the G1’s cameras, LiDAR, joints, and IMU to write and run programs.
[HT]
