AI Coding Agents Autonomously Program and Train Physical Robots Without Human Supervision

Two distinct systems for autonomous robot training emerged in the same week in June 2026. NVIDIA's ENPIRE framework, built with Carnegie Mellon University and UC Berkeley, uses AI coding agents to run robot training experiments overnight without human supervision — demonstrating robotic arms learning to cut zip ties and install GPUs into motherboards [1]. Separately, Anthropic's Project Fetch Phase 2 showed Claude Opus 4.7 independently programming a robot dog in 12 minutes and 7 seconds, approximately 20x faster than a human-assisted team completed the same task in the prior year [3][2].

If AI agents can autonomously run the trial-and-error loops that robot training requires — writing code, interpreting sensor feedback, adjusting policies, and repeating — the pace at which robots acquire new skills could increase without proportional human labor. Both NVIDIA and Anthropic demonstrate this on real hardware, not simulation.

NVIDIA's GEAR lab, working with Carnegie Mellon University and UC Berkeley, released ENPIRE — a framework that wraps AI coding agents with memory, context management, constraint handling, and feedback loops to run robot training experiments autonomously [1]. Researchers set tasks, then let the system operate overnight: by morning, a report describes what the agents tried and how robot performance changed. The system has been used to train robotic arms on dexterous manipulation including cutting zip ties and inserting GPUs into motherboard sockets. NVIDIA's own framing of the result: 'A part of our NVIDIA GEAR lab now self-improves tirelessly overnight. We just read the reports in the morning' [1].

On the same day this research attracted broad attention, Anthropic published results from Project Fetch Phase 2 [2]. In the original 2024 experiment, Project Fetch asked human employees — aided by Claude — to program an off-the-shelf robot dog from scratch. In Phase 2, Claude Opus 4.7 was given the task alone: connect real robot hardware, read camera and lidar sensor feeds, write movement code, and track the robot's location. The model completed the full sequence in 12 minutes and 7 seconds [3], roughly 20x faster than the human-assisted team in the prior year [3][4].

A parallel body of academic work covers the methods that underlie both systems: language-instructed skill acquisition, continual robot learning, and LLM-guided reinforcement learning for real-world manipulation [5][6][7][8]. These papers situate ENPIRE and Project Fetch within wider research on using LLMs to reduce the human effort in robot learning pipelines. The NVIDIA and Anthropic work is distinctive for running against real hardware rather than simulation, and for framing the result as a production capability rather than an ablation study.

One notable gap in the coverage: a social media post asserts that every major AI lab except OpenAI and Anthropic is building physical AI [9]. Project Fetch Phase 2 directly contradicts the Anthropic half of this claim, but neither company has announced a sustained robotics research division comparable to efforts at Google DeepMind or dedicated robotics firms. Whether these experiments represent ongoing strategic bets or periodic capability probes is not addressed in the available sources.