TOOL

uni-robot-agent (184 stars): one free toolkit to test whether an AI agent can run a robot before you buy one.

uni-robot-agent is an MIT-licensed harness that runs robot-use agents on Codex or Claude Code, tests them in the RoboDojo simulator, and can hand motion to a Pi0.5 action model. Tasks are lab work like stacking bowls, not plant work. It has 184 GitHub stars.

For operations and automation teams: test whether an agent can drive a robot before buying one.

More from The Feed

NEWS

ADNOC runs field master planning in hours, not months, as the first energy company on OpenAI Frontier.

ADNOC signed a multi-year deal to run OpenAI Frontier on G42's UAE infrastructure, the first energy company to deploy it. Named uses are field master planning in hours instead of months and agents that value every barrel across production, refining, logistics and trading. ADNOC says human experts keep oversight, which is the line to watch.

ADNOC signs partnership agreement with OpenAI and G42
Image: ADNOC (via Gulf News)
NEWS

CATL's Spirit AI robot inserts wires with over 99% success as China's robot makers race to build robot brains.

Reuters reports Chinese embodied-AI firms raised about $13B across 322 deals in the first half of 2026. CATL says Spirit AI robots insert flexible wires in battery-pack testing at over 99% success, and DexForce is readying some factory tasks in 6 to 8 hours. The factory floor is where this money is now being judged.

Found via Mango (@Mango_Wg) on X