# FieldProof by 180Climate - **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:** [Kian Boon TENG](https://cerebralvalley.ai/u/johnteng) - **GitHub:** https://github.com/TengKianBoon/fieldproof - **Demo video:** https://youtu.be/pFf991QwpUU - **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery - **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1312 AI can't walk into a forest. People can. FieldProof lets a client, or the client's AI agent, hire a local professional to collect evidence on the ground, and makes that evidence easy to check. Built on Meta: Muse Spark 1.3, through the Meta Model API, is our agent, label reader, job drafter and review assistant. Muse Code is the coding agent we built it with during the hackathon. Why it matters Our teams in Indonesia collect samples by boat and forest track. Field work still ends with photos in a chat group and paperwork that is hard to check. One blurry photo or wrong label can mean repeating the whole trip. Mining and property due diligence face the same problem. How it works (live) 1. Hire a human. A client posts a field job with a budget, or their Muse Spark agent drafts one through our MCP server. A person publishes it, invites local professionals and picks a bid. 2. Guided capture. The professional opens a phone link, with nothing to install, and is guided item by item. They never see the expected answers. 3. Optional location check-in, shown on a map beside the planned site with distance and accuracy. 4. Muse Spark reads every label as structured output: the code, and whether the label was clear, blurred, covered or missing. It leaves the code empty rather than guess. Our server flags mismatches such as B71 where B17 was expected, and never auto-corrects. 5. A person decides every item, with a plain-English brief from Muse Spark on request. 6. The evidence comes back: a printable report with photo fingerprints, a JSON evidence pack and a follow-up check for flagged items. The client's agent can collect it read-only; it can't publish, choose or decide. Real trial 9 people ran an assisted home trial on their own phones with handwritten tags. In the submissions we checked, FieldProof caught the planned B17/B71 mismatch, and 8 sent feedback that shaped our next steps. Built to hold up Signed, expiring links per person, a daily ceiling on model spend, protection against duplicate requests, clear error screens, and 377 automated tests passing with a strict TypeScript check. What's next - Meta Ray-Ban Display glasses: the checklist in view, hands-free photos, and live voice guidance with instant translation. Try the 600 x 600 glasses view today (change /work/ to /glasses/ in any field link). - Pilot jobs in the 180Climate Fieldwork network, and location-based hiring with distance and cost estimates. - FieldProof inside Meta AI through our MCP tools, capture without signal, and small on-device Llama models. Business model: a per-job evidence fee plus an introduction fee on accepted jobs. Try it: https://fieldproof-180climate.netlify.app (judge quick-start in the README) · Judge access code: Beautiful123$ Video: https://youtu.be/pFf991QwpUU Code: https://github.com/TengKianBoon/fieldproof FieldProof is the evidence layer of 180Climate's Fieldwork network. All FieldProof code was written during the hackathon, from 27 October. ## More from Meta Global AI Developer Hackathon - [Hydrosight](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1309) - [Tin Chu](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1310) - [Crucible](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1311) - [OPIE CHISOM SAMUEL](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1313) - [Shunfeng](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1314) - [Omni-Ouro Inc.](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1315) --- Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1312. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt