# Robot-Agent-Lab

- **Event:** [The Recursive Agents Hackathon](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon)
- **When:** Sat, Oct 17 at 9:00 AM – 10:00 PM (PDT)
- **Where:** San Francisco, CA
- **Team:** [Abhijit Betigeri](https://cerebralvalley.ai/u/abhijitbetigeri)
- **GitHub:** https://github.com/abhijitbetigeri/robot-agent-lab
- **Demo video:** https://abhijitbetigeri.github.io/robot-agent-lab/assets/demo.mp4
- **Gallery:** https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/24

Robot Agent Lab: a team of Nemotron agents runs a science lab's night shift on a simulated SO-101 arm. Tubes arrive with coloured caps. Red goes to the cold rack, blue to the sequencer rack, green stays on the bench. A tube in the wrong rack ruins an expensive run. The human types "Put tonight's samples where they belong." and leaves.

Five agents, one loop, all on Crusoe. A planner writes checks that can be decided from a photo. An operator drives the arm and must cite evidence for every check before claiming done. A sealed verifier sees only the checks and one photo, never the operator. A monitor catches loops and stalls and decides to continue, re-plan, stop or ping a human. A curator writes each shift into a shared memory ledger where a new value supersedes the old one and nothing is deleted. An improver rewrites the team's playbook and keeps it only if it scores better on a shift it has never seen.

The bench has quirks the memory must learn: a gripper that ignores the first close, a rack that moves, a nudged camera. The simulator gives ground truth, so we measure the verifier's own accuracy. In the recorded runs, every false "done" was caught, the moved rack was superseded with zero stale positions served, and the first playbook rewrite was kept. Memory has not yet reduced robot actions; the README says so.

## More from The Recursive Agents Hackathon

- [Ming](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/21)
- [Shadowbox](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/22)
- [Concord](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/23)
- [Done](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/25)
- [AYA](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/26)
- [TicketZero](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/27)

---

Markdown version of https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/24. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt