AI coding agents autonomously train robots to install GPUs and cut zip ties

Researchers at Nvidia's GEAR lab, in collaboration with Carnegie Mellon University and UC Berkeley, developed ENPIRE, a software framework that enables AI models to autonomously direct robot training. Using a generous token budget, the AI agents successfully taught robotic arms to cut zip ties and insert GPUs into motherboard sockets. The system functions fully autonomously, allowing the lab to improve overnight with human researchers reviewing results in the morning.