BugBuster
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Blackbox helps people preserve and examine video after a road crash. A user opens a case, uploads up to three clips, and receives a review workspace that keeps the original files traceable through SHA-256 hashes. The system indexes video frames, proposes object tracks, links observations to source moments, identifies missing evidence, and exports a PDF report and structured JSON package. The central design choice is to show what the footage supports and what it cannot establish. A candidate track requires human review. Physical speed requires suitable calibration and scale. Simulated views are labeled illustrative, and Blackbox does not assign legal fault. Our demo uses synthetic crash footage. It shows a prepared case with one indexed clip, candidate tracks generated with Meta SAM 3.1, a source-linked observation, an evidence checklist, and traceable exports. Blackbox is a prototype intended to give people and qualified investigators a more accessible starting point for evidence review. Independent accuracy validation and real-world pilots are still needed.