# HOLDOUT

- **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:** [hima vanth](https://cerebralvalley.ai/u/himavanth)
- **GitHub:** https://github.com/himavanthkar/proof-of-done/tree/main
- **Demo video:** https://github.com/himavanthkar/proof-of-done/blob/main/media/proof_of_done_demo.mp4
- **Gallery:** https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/17

AI agents fill vague goals with their own assumptions, then confidently report "done". Proof of Done is an agent team on Crusoe Managed Inference that is not allowed to say "done" without evidence.

From a one-line goal:

A spec agent writes explicit assumptions and pytest acceptance tests, locked by SHA-256 so the builder can't edit them.
A builder writes the code in a sandbox.
Mutation testing deliberately breaks the code to prove the tests are strong; weak tests get strengthened and re-locked.
An NVIDIA Nemotron verifier audits the goal, assumptions, tests, code and evidence, and can dispute a test it believes is wrong.
A deterministic Python rule allows DONE only when the tests pass, at least 80% of mutants are killed, and the verifier approves. Otherwise it retries, re-plans, fixes a disputed test once, or asks the human one specific question instead of guessing.
Every run ends with a receipt (assumptions, test hash history, mutation results, verdict, timeline, cost).

## More from The Recursive Agents Hackathon

- [Repro](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/14)
- [CORE GENESIS](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/15)
- [TeamAIWork](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/16)
- [AutoFab](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/18)
- [Crusoe-Cooler-System](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/19)
- [Nemotron Done Means Done](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/20)

---

Markdown version of https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/17. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt