# WireProof

- **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:** [Godswill Idolor](https://cerebralvalley.ai/u/Gwill)
- **GitHub:** https://github.com/big14way/wireproof
- **Demo video:** https://youtu.be/TkUPQL70WRc
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/866

WireProof is a hands-free copilot for the electrician wiring a Nigerian house, and proof for the homeowner, before the walls are plastered.

In Nigeria we build with block and plaster. The wires go into the wall, the plaster goes on, and from that day nobody can see them again. Electrical faults are the leading cause of the fires Lagos Fire Service counts every year: fake copper cable, cable too thin for the breaker in front of it, boards with no RCD. All of it is hidden before anyone checks.

The electrician opens WireProof on the phone already in their pocket, starts a job (house, owner's WhatsApp number, circuits on the board), and talks to it while they work. Muse Voice Transcribe hears Nigerian English. Muse Spark reads breaker boards and cable prints off the sheath, plans each step as the agent, and asks the electrician to check by eye when a print is ambiguous. SAM 3.1 traces every conduit run in the wall using the blocks as a ruler, and Muse Image shows the owner the wall after plaster with the cables drawn where they really are. Muse Spark Open runs a privacy gate so no photo with a face in it leaves the phone.

The model reads; the code decides. Every reading is verified by deterministic rules citing BS 7671 (the size must appear in the text the model copied, a rating must match the device, every number in a reply must come from a tool or the electrician) or it is refused and the electrician is asked to reshoot. On 21 real cable photos it read 20 right and caught the one it got wrong; on 3 real boards it read 18 of 20 devices and found no RCD on any of them. When the electrician signs off, the owner gets a screening report on WhatsApp: the verdict, the fixes in order, and the photo behind every one.

A working electrician tried it on their own phone (that is the phone footage in the video). Their feedback, that the copilot's words should be more like Nigerian Pidgin for those who don't follow English well, is now a "Talk Pidgin" switch: Muse Spark renders the verified reply in Pidgin, and code checks it states exactly the same numbers before it is spoken.

Live site: https://wireproof-production.up.railway.app (the public panel and the demo report need no passcode; the copilot and console are behind a passcode shared with judges on request). Screening, not certification: a licensed electrician signs off every job.

## More from Meta Global AI Developer Hackathon

- [Storey](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/863)
- [Valorm Technologies](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/864)
- [Areazero.ai](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/865)
- [StarAppsPh](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/867)
- [Shubhadeep Personal](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/868)
- [NookShift - Indie](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/869)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/866. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt