# Témoin

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [EPHISCOWELL OUAKEU KENMEUGNE](https://cerebralvalley.ai/u/EPHISCOWELL)
- **GitHub:** https://github.com/mysteriolaflamme/Temoin
- **Demo video:** https://youtu.be/ydtwjkeiSR4
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2870

In Africa, most trade is informal: items are sold on WhatsApp and Facebook, paid at delivery, and disputes are settled word against word. In Europe, return fraud costs honest sellers dearly. Témoin brings a neutral witness to every remote second-hand sale.

A used item is unique because of its flaws, like a fingerprint. At the sale, Témoin maps every flaw and issues a one-time challenge code the seller writes next to the item, so old photos cannot be reused. At delivery, the buyer gets a new code and photographs the item; an investigator agent then runs a multi-step enquiry, zooming on doubtful flaws, checking both codes and matching flaw by flaw, and deterministic rules decide: MATCH, DAMAGED IN TRANSIT, NEEDS REVIEW or MISMATCH. Wrongful accusation is treated as the worst error, so doubt always means "needs review".

Every verdict is sealed in a SHA-256 hash chain with its full investigation trail, verifiable by anyone. Users can talk to the case by voice and get answers grounded only in the sealed evidence, read aloud by a Meta voice on a plain CPU. A business REST API lets any marketplace or courier embed Témoin in four calls.

Evaluation: 90% accuracy and zero false accusations on our test pairs. Témoin assists a human decision; it never replaces it.

## More from Meta Global AI Developer Hackathon

- [ParkSpot-AI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2867)
- [MetaStudy](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2868)
- [London Pipeline](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2869)
- [Quietlog](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2871)
- [SpatialSentry](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2872)
- [Lunes](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2873)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2870. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt