War Rooms — live incident drills
Investigate a simulated production incident, resolve it, and get graded like an SRE.
- Practice real incident response — triage, RCA, mitigation — without breaking anything real
- The AI channel only reveals evidence you ask for, so you learn to investigate, not guess
- A solid resolution earns XP and can raise your mapped competency level
- Study the Failure Library of real-world incidents behind many drills
You're paged. Go.
A War Room drops you into a live production incident: a page, a situation, and a severity. Your job is the on-call engineer's job — triage, investigate, mitigate, prevent, write it up.
Investigate through the incident channel
The left panel is your incident channel and on-call tooling in one. Ask it for specific evidence — "show the error logs", "what deployed in the last hour?", "database metrics?" — and it replies like a monitoring tool pasting output. It reveals only what you ask for and never hands you the root cause. If you propose an action (rollback, restart, scale), it describes the observable effect without confirming you're right — just like production.
Resolve and debrief
When you think you have it, fill in the five resolution fields:
1. Triage & impact — severity, who's affected, mitigate vs investigate
2. Root cause — what happened, with the evidence that proves it
3. Immediate fix — the mitigation now, with rollback awareness
4. Durable prevention — what stops the recurrence
5. Post-mortem — blameless: timeline, contributing factors, action items
You need at least triage, root cause and fix before you can declare resolved. An AI grader — playing a principal SRE running a blameless debrief — scores you /100 against the drill's rubric, shows a per-dimension breakdown, tells you whether you found the true root cause, and summarises how the best engineer would have handled it.
What you earn
Score 60+ and you earn XP (up to 25, proportional to score) and a streak bump. Score 85+ and the drill's mapped competency can rise to verified level 3. Your best score per drill shows on the hub.
The Failure Library
The second tab collects real incidents from industry. Drills tagged based on a real incident link to the real story after your debrief — read it and compare notes with history.
The journey, step by step
| Step | What you do | What you get |
|---|---|---|
| Pick a drill | Open War Rooms and choose an incident from the Live Drills tab | You're paged: a brief, the situation, and a severity (major / severe / catastrophic) |
| Investigate | Ask the incident channel for logs, metrics, recent deploys, DB state | Evidence appears only when you request it — like real on-call tooling |
| Resolve | Fill in triage, root cause, fix, prevention and post-mortem | At least triage, root cause and fix are required to declare resolved |
| Debrief | Declare resolved & debrief | AI grades you /100 with a dimension breakdown, feedback, and how the best would have handled it |
| Bank it | Score 60+ on the debrief | XP awarded (up to 25), streak bumped, and the incident's competency can level up |
Frequently asked
Why won't the channel just tell me what's wrong?
By design. The AI plays your monitoring stack and teammates: it reveals only the evidence you ask for and never volunteers the root cause. Diagnosis through evidence is the skill being trained.
How is my incident handling scored?
An AI grader acting as a principal SRE scores you /100 against the drill's rubric — comparing your root cause, fix and prevention to the ideal ones — and tells you whether you found the true root cause.
Do war rooms earn XP?
Yes — scoring 60 or higher earns XP proportional to your score (up to 25), bumps your streak, and a strong run (85+) can raise the verified level of the drill's mapped competency.
What's the Failure Library tab?
A catalog of real-world incident case studies. Drills marked "based on a real incident" link back to the real story after your debrief, so you can compare your handling to history.
Related guides
Pitch, de-escalate, negotiate and present against an AI that stays in character — then get scored.
ReadLook at a real screen and reconstruct the spec that produced it — graded instantly by AI.
ReadHow the in-browser runner and tests work.
ReadHow the XP and spoken-skills leaderboards rank you, and what the dashboard league resets.
Read