The system uses AI coding agents to train fleets of physical robots with a 99% success rate on precision tasks, no humans required.
Nvidia just built a system that lets AI agents walk into a robot lab and figure out how to teach the robots themselves. No human hand-holding, no step-by-step programming. Just coding agents, robotic arms, and what the researchers describe as a “generous token budget.”
The framework is called ENPIRE, short for Environment, Policy Improvement, Rollout, Evolution. It was developed by Nvidia’s GEAR (Generalist Embodied Agent Research) lab in collaboration with Carnegie Mellon University and UC Berkeley. The result: robots that learned to cut zip ties and insert GPUs into thin motherboard sockets with a 99% success rate.
How ENPIRE actually works #
The framework has four distinct modules. The Environment module handles automated resets and verification, so a robot can fail at a task and immediately set itself up to try again without someone walking over to press a button. Policy Improvement is where the agent analyzes performance data and adjusts the robot’s behavioral code. Rollout handles parallel physical evaluation across multiple robots simultaneously. And Evolution is the agent-driven code refinement layer that ties the whole cycle together.
The research team deployed eight dual-arm robots running parallel policy rollouts. The researchers tracked efficiency using metrics they call Mean Robot Utilization (MRU) and Mean Token Utilization (MTU), essentially measuring how much useful work each robot and each AI token produced.
The coding agents powering the system included Codex via GPT-5.5, Claude Code, and Kimi Code.
Why this matters beyond the lab #
Nvidia’s Jim Fan characterized the achievement as enabling “AutoResearch in the physical world for the first time.” Most AI breakthroughs live in software, in language models generating text or image models generating pixels. ENPIRE operates on actual hardware. Physical objects. Real-world physics with all its messy unpredictability.
The project builds on previous GEAR lab work including GR00T humanoid models and DreamGen synthetic data generation. ENPIRE adds a closed-loop system where agents can improve robot performance directly on physical hardware without continuous human oversight.
Nvidia announced the framework during GTC 2026 and has indicated plans to open-source it, which would allow external developers and companies to set up their own self-running robotic labs.
What this means for investors #
This announcement had zero connection to crypto, blockchain, or decentralized networks. No token launches, no on-chain integration, no Web3 buzzwords.
Every robot in the ENPIRE setup runs on Nvidia hardware, uses Nvidia’s software stack, and feeds data back through Nvidia’s compute infrastructure. If autonomous robot training scales the way this research suggests it could, the demand for Nvidia’s GPU and robotics platforms grows proportionally.
The open-source angle introduces a wildcard. If ENPIRE becomes freely available, it lowers the barrier to entry for smaller robotics startups, potentially fragmenting a market that has been dominated by well-capitalized players.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our