AI can write the code. War Rooms measure whether you can command it when production is on fire: drop into a live incident and drive the decisions AI can't make — triage, RCA, safe fix, prevention and post-mortem — graded like a staff SRE and logged to your auditable record.
Not a quiz about incidents — the full loop a senior engineer drives when production is on fire, and five clean tests of the judgment AI can't hand you.
Assess severity and blast radius, decide mitigate-vs-investigate, and page the right people — under a ticking clock. The first judgment call AI can't make for you.
Pull logs, metrics, recent deploys, traces and config from an AI incident channel that reveals only what you ask for — it never hands you the answer. The discernment is in what you choose to ask.
Choose and justify the fix that stops the bleeding without making it worse — rollback, flag, hotfix or scale — and defend the trade-off. Command the change; own the call.
Design the durable change so this class of failure can't recur: the missing guardrail, test, alert or circuit breaker.
Write the timeline, contributing factors and action items the way a healthy software org actually does.
Investigation you have to drive, grading that holds a real bar, and incidents drawn from history — prime evidence in your auditable record and toward the Itz'at.
A realistic on-call channel role-plays the incident, revealing evidence only when you ask the right question — so you practise investigation, not guessing. AI surfaces signals; you decide what they mean.
Scored across triage, investigation method, correct root cause, safe fix, durable prevention and communication — with specific feedback that lands in your auditable record.
Drills span classic failure modes and real documented outages (GitLab, Knight Capital, Cloudflare, S3) with variations to reason through.
Every drill is paired with a browsable library of real documented failures — Knight Capital's $440M/45min, the AWS S3 typo, Cloudflare's regex CPU meltdown, Equifax — each mapped to the principle it teaches and the cost of ignoring it. Every call you make feeds your auditable record and counts toward the Itz'at — miatz's proof-of-judgment credential.
Incident scores enter your auditable record and count toward the Itz'at and mastery-gated advancement.
Each failure teaches a concrete concept — idempotency, circuit breakers, rate limits, rollback.