AI Coding Agents Autonomously Teach Robots to Install GPUs and Cut Zip Ties

Researchers at NVIDIA’s GEAR lab, in collaboration with Carnegie Mellon University and UC Berkeley, have developed ENPIRE, a framework that lets AI coding agents autonomously train robots. The agents used a generous token budget to teach robots tasks such as cutting zip ties and inserting GPUs into motherboard sockets. According to NVIDIA AI director Jim Fan, the system now self-improves overnight with researchers reviewing the results each morning.