# ProofPath

- **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:** [Pankaj Verma](https://cerebralvalley.ai/u/Pankajverma)
- **GitHub:** https://github.com/consultanything-tech/ProofPath
- **Demo video:** https://storage.googleapis.com/proofpath-hackathon-2026-submission-media/proofpath-demo.mp4
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/727

ProofPath helps developers turn checkout acceptance criteria into reviewed browser tests and inspectable regression evidence. Muse Spark 1.3 proposes bounded semantic test cases grounded in registered checkout controls. A person reviews the interpretations, reference policies and coverage gaps, then approves an exact snapshot before Playwright can execute it. An independent typed oracle computes expected results; the model cannot declare PASS or override browser evidence.

The workbench compares a previous and candidate release, links assertions to screenshot hashes and observations, and exports a bug packet or Playwright regression. Reviewed definitions can be reused against a fixed release without another planner call. Candidate PASS and overall suite INCONCLUSIVE remain separate when coverage is incomplete.

Live product: https://proofpath-web-vgrq2gffta-as.a.run.app/workbench.html
Reviewer access code: bK-uaqfvVNBT6KOROea-Mqt0UwlKm10B
Open the live product, enter the code and select Unlock Meta testing. Use the prefilled synthetic checkout requirement, review the proposed checks, approve the plan and run. The existing shared proposal limits apply. Viewing saved reports does not spend model allowance.

Built during this hackathon; pre-event work was planning documentation only. The narrated video uses the actual local workbench with explicitly disclosed fixture plans and synthetic data. Separate committed evidence records a real Meta-generated V2 plan, a seeded regression and a saved fixed rerun. GCP hosting, authenticated saved reports and evidence downloads were verified; a new model generation on the hosted service has not been smoke-tested.

## More from Meta Global AI Developer Hackathon

- [RAJAKUMARAN](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/724)
- [Point, Say, Press](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/725)
- [Agentic Verification Layer](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/726)
- [Farhan](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/728)
- [EcoAgents](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/729)
- [Muhammed Taha Aktaş](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/730)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/727. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt