AI coding agents taught robots how to install GPUs and cut zip ties

2026-06-17 · Source: AI - Ars Technica · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Software Development & Engineering · Depth: Advanced, short

Summary

Nvidia's GEAR lab, in collaboration with Carnegie Mellon and UC Berkeley, developed ENPIRE, an agent harness framework enabling AI coding agents to autonomously train robots. This framework, detailed in a June 16, 2026 research paper, allows agents to perform automatic task reset, refine robotic policies, evaluate them across multiple physical robots, and address failures by analyzing logs and improving code. Tested with OpenAI's Codex (GPT-5.5), Anthropic's Claude Code (Opus 4.7), and Moonshot AI's Kimi Code (Kimi K2.6), the system achieved a 99 percent success rate on manipulation tasks like cutting zip ties and inserting GPUs. Eight-agent teams reached 99 percent success on the Push-T task in two hours, outperforming smaller teams. However, limitations include robot idle time during agent processing, increased token consumption for larger teams, and occasional underutilization of compute resources. Nvidia is also expanding its physical AI vision through partnerships with Unitree and Hyundai.

Key takeaway

For Robotics Engineers aiming to accelerate robot skill acquisition, autonomous AI coding agents leveraging frameworks like ENPIRE offer a path to significantly faster training cycles. You should evaluate integrating such agentic systems into your lab, recognizing that while larger agent teams achieve quicker success, careful management of compute resources and token consumption is crucial to optimize efficiency and control operational costs.

Key insights

An agentic harness allows AI coding agents to autonomously train robots, achieving high success rates in complex manipulation tasks.

Principles

• Agentic harnesses enable autonomous AI control.
• Parallel robot evaluation accelerates training.
• Larger agent teams achieve faster success.

Method

The ENPIRE harness employs modules for automatic reset/verification, policy refinement, parallel evaluation, and failure analysis through logs and code improvement.

In practice

• Deploy multi-agent teams for faster skill acquisition.
• Integrate feedback loops for autonomous policy refinement.
• Utilize agentic frameworks for self-improving robot labs.

Topics

• AI Coding Agents
• Robotics Training
• ENPIRE Framework
• Autonomous Systems
• Manipulation Tasks
• Token Economy

Best for: Research Scientist, AI Scientist, Robotics Engineer, Machine Learning Engineer

Related on AIssential

• Nvidia research shows robots that train themselves through AI coding agents — AI coding agents empower robots to autonomously achieve up to 99% success in real-world dexterous grasping tasks.
• not much happened today — AI development is rapidly advancing across agentic systems, inference efficiency, and model architectures, while grappling with practical deployment and ethical challenges.
• Everyone Will Be Training AI Agents This Way in 6 Months (Claude Code) — Iterative training with feedback enables AI agents to autonomously develop and refine complex, multi-step skills.
• Autoresearch, Agent Loops and the Future of Work — AI agentic loops enable autonomous, iterative experimentation and continuous improvement across various domains.
• The Month AI Woke Up — AI agents have matured to enable autonomous task execution, fundamentally altering programming and impacting market valuations.
• How To Train Your AI Agent's Skills (Claude Code / OpenClaw) — Iterative skill training for AI agents using markdown files drastically improves efficiency and reduces operational costs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI - Ars Technica.