From idea to working war room in one weekend
War Room is submitted! A quick log of how it came together:
Day 1 — Make it real. Our first fault injector just wrote a fake alert — and the
"investigation" was rigged. We threw it away and rebuilt: now "deploying an incident"
literally git checkouts a guilty commit into a live service, real traffic fails,
and the alert fires from an error-rate threshold over actual logs.
Day 2 — The agent team. Five GPT-5.6 agents now run the full loop with their reasoning streamed live: Triage → Log Detective → Git Detective (names the guilty SHA from a multi-commit evidence window) → Fixer (commits a fix branch, redeploys, and measures recovery: error rate 100% → ~5% on screen) → Scribe (writes the postmortem).
Day 3 — Judge-proofing. Cold-cloned the repo on a clean machine: two commands, zero credentials, full incident loop working. A keyless demo mode means anyone can try it without an API key — with a key, all five agents run live on GPT-5.6.
Try it:
Log in or sign up for Devpost to join the conversation.