Nvidia ENPIRE lets robots master physical tasks on their own
- Source
- Ars Technica
- Time
- 3:27 AM
- Weight
- 95/100
Nvidia researchers, in collaboration with teams from Carnegie Mellon University and UC Berkeley, have developed an autonomous "agent harness" framework called ENPIRE. This software allows AI coding agents to autonomously train robots to perform complex physical tasks, such as inserting GPUs into motherboards and cutting zip ties.
By providing agents with memory, context, and feedback loops, the system enables them to independently develop training strategies, analyze logs, and refine robotic behaviors without human supervision. Testing revealed that teams of multiple AI agents, using models like GPT-5.5 and Claude 4.7, could achieve success rates of up to 99 percent, often outpacing human-led training methods.