NVIDIA GEAR lab researchers alongside collaborators from Carnegie Mellon University and UC Berkeley developed ENPIRE, an agent harness framework that enables AI coding agents to autonomously direct robot training, according to a research paper uploaded on June 16, 2026. The framework successfully trained robots to perform tasks including cutting zip ties and inserting GPUs into motherboard sockets. Jim Fan, director of AI at NVIDIA, stated in a LinkedIn post that part of the NVIDIA GEAR lab now self-improves overnight with researchers reviewing reports in the morning.
ENPIRE Framework Enables Autonomous Robot Training
ENPIRE is an agent harness framework that wraps around AI models to enable their use of various tools while providing capabilities such as memory, context, constraint, and feedback loops. The framework was developed by robotics researchers at the NVIDIA GEAR (Generalist Embodied Agent Research) lab. AI coding agents using the framework were given a lab full of robotic arms, compute resources, and a token budget for teaching robots various tasks.
Four-Module Architecture Supports AI Agent Operations
The ENPIRE harness has four modules that enable AI coding agents to perform automatic reset and verification on tasks, refine policies that guide robotic behavior, evaluate such policies across multiple physical robots working in parallel, and address failures by analyzing logs, ingesting research papers, and improving training infrastructure and algorithm code. More technical details are available in the research paper uploaded on June 16, 2026.
Three AI Coding Agents Tested Across Multiple Robots
The harness was tested with three different AI coding agents: OpenAI's Codex with GPT-5.5, Anthropic's Claude Code with Opus 4.7, and Moonshot AI's Kimi Code with Kimi K2.6. Teams of the coding agents independently developed different algorithmic approaches to robot training, tested them in real-world experiments, and then retained whatever changes helped raise the overall success rate over repeated cycles of self-directed testing.
NVIDIA Plans Open-Source Release of Framework
Jim Fan stated the team would be open-sourcing everything so anyone can host their own self-running robot lab at home. Fan also described the goal of such AI-directed robot training, saying researchers could take a holiday and NVIDIA founder and CEO Jensen Huang would not notice.
FAQ
What is ENPIRE and who developed it?
ENPIRE is an agent harness framework developed by NVIDIA GEAR lab researchers alongside collaborators from Carnegie Mellon University and UC Berkeley. The framework enables AI coding agents to autonomously direct robot training by wrapping around AI models to provide capabilities such as memory, context, constraint, and feedback loops.
What tasks did AI agents successfully train robots to perform using ENPIRE?
AI coding agents using the ENPIRE framework successfully trained robots to cut zip ties and insert GPUs into thin sockets on motherboards. The agents figured out a training regimen when given a lab full of robotic arms, compute resources, and a token budget for teaching robots various tasks.
