@DrJimFan: Today, we enable AutoResearch in the physical world for the first time! Introducing ENPIRE: we give 8 Codex agents a fl…
Summary
NVIDIA GEAR lab introduces ENPIRE, a system that uses 8 Codex agents to autonomously control a robot fleet for physical tasks like tying zip-ties and installing GPUs, demonstrating self-improving robotics research and a new 'physical scaling' phenomenon.
View Cached Full Text
Cached at: 06/16/26, 07:39 PM
Today, we enable AutoResearch in the physical world for the first time! Introducing ENPIRE: we give 8 Codex agents a fleet of robots, an allocation of GPUs, and generous token budget. We set them free with a simple goal: solve the task as quickly as possible, keep the robots busy but stay safe, don’t waste precious compute. Make no mistake.
Then humans step aside and our watch begins. The robot fleet starts to come alive: they learn to look for visual clues, reset the scene, practice novel skills, tinker with control stack, read papers online, debate, reflect, get stuck, and try again directly on the hardware. All we did is to give Codex an API to the world of atoms, and the rest is emergence.
ENPIRE is able to solve high-precision tasks like tying zip-ties, organizing fine pins, and installing GPUs all by itself. We also discovered a new type of “physical scaling”: 8 robots exploring in parallel improves significantly faster than fewer ones.
A part of our NVIDIA GEAR lab now self-improves tirelessly over night. We just read the reports in the morning.
/goal: we all take a holiday and Jensen wouldn’t even notice ;)
We will be open-sourcing everything, so you can host your self-running robot lab at home too! Deep dive in the thread:
Project site: https://research.nvidia.com/labs/gear/enpire/…
Wenli has written an excellent technical thread, please check it out!
And 10x more exciting. Design space for the harness is also huge
We set up and verify the safety measures on the robots manually. That has to be set up before autorsearch happens
A lot more incoming!
Exciting progress towards true ASI for science! @DrJimFan
Physical Superintelligence!!
Ikr! Robots humming tirelessly at night
Hey @DrJimFan, you might like what we did last Saturday in 8 hours!
@DrJimFan your agents get a clear reward signal - did the robot grasp the object. I run 200+ daily actions at http://monday.com where the output is a judgment call with no ground truth. physical AI gets a loss function from physics. who writes one for an agent whose output is an opinion?
Similar Articles
@FinanceYF5: Jim Fan announces that the team has brought AutoResearch into the physical world for the first time, launching ENPIRE. They equipped 8 Codex agents with robots, GPUs, and sufficient tokens, allowing them to autonomously learn, trial and error, collaborate, and complete tasks on real hardware.
Jim Fan announced that the team launched ENPIRE, bringing AutoResearch into the physical world for the first time, equipping 8 Codex agents with robots, GPUs, and tokens, enabling them to autonomously learn and collaborate on real hardware to complete tasks.
AI coding agents can autonomously direct robot training
AI coding agents using the open-source ENPIRE framework can autonomously train robots to perform tasks like installing GPUs and cutting zip-ties, with the system self-improving overnight.
How NVIDIA engineers and researchers build with Codex
NVIDIA engineers and researchers are using OpenAI Codex, powered by GPT-5.5, as their default tool for complex engineering tasks and end-to-end machine learning workflows. The article highlights significant productivity gains, autonomous system building, and research automation achieved through this integration on NVIDIA infrastructure.
NVIDIA Enables the Next Era Of Physical AI Research With Agent Skills For Autonomous Vehicles, Robotics And Vision AI
NVIDIA announced new physical AI agent skills at CVPR to accelerate research in autonomous vehicles, robotics, and vision AI, including tools for neural reconstruction, simulation, and reinforcement learning.
NVIDIA Research Unlocks Advanced Grasping, Smarter Autonomous Driving and Agent Training at Scale
NVIDIA Research presents three papers at CVPR: GraspGen-X (zero-shot grasping foundation model), LCDrive (efficient autonomous driving reasoning), and NitroGen (generalized gameplay AI foundation model), highlighting training at scale for physical AI systems.