On-call practice
Practice your first production outage before you're responsible for a real one.
You're on call. Checkout is failing. Investigate with real-looking metrics, logs and deploys, choose a fix, and watch the simulated service respond. Then replay the same outage with one thing changed.
Investigate
Every look takes a simulated minute. Customers keep failing while you work.
Act
Rollbacks, flags and config changes take time to roll out, and can make things worse.
Debrief
Every point in the debrief links to the exact step in your timeline that it's based on.
The outcome of every action is decided by a fixed simulation, not by an AI model, so the same choices always lead to the same result.
Judging this project? Enter your judge code
Investigate
Each look takes 1 simulated minuteAct
Changes take time to roll outPost a status update
Teammate
AI (Nemotron) · coaches, never decides outcomesTimeline
Recorded exactly as it happenedDebrief
What happened
Facts from the simulation · each one cites your timelineYour timeline
Coach's read
AI interpretation (Nemotron). Citation IDs and numeric values checked; reasoning may still be wrong.Original run vs. replay
From recorded stats only| The page | Rollback unsafe |
|---|
Stress-test your plan
The simulator decides every result. Nemotron only translates and explains.Write what you would do the next time this page fires. Nemotron turns your words into simulator steps, keeping your plan as written, even its mistakes. You check the steps, then the simulator runs them on two versions of this outage: A, where the release's database migration can be undone, and B, where it cannot.
Pressure Replay
Same outage. This time, rollback is unsafe.
Release v2.14.0 shipped an irreversible database migration. Rolling back makes almost every checkout fail. Find another way.