Reconstruct the whole run from one correlation ID

The prompt

Correlate model, tool, policy, CI, approval, deployment, latency, cost, and outcome data under a single execution ID. Confirm an operator who wasn't involved in the run can reconstruct exactly what happened and compare it against your human-only baseline.

Why it's built that way

Exit criterion: an operator can reconstruct a run and compare it with the baseline. This is also the capstone's evidence package and final-demonstration bar — the demo is successful only if the controls are visible, not merely if the code works.

What to do with the answer

Hand the evidence package to someone who wasn't in the room for the run and ask them to explain what happened — if they can't, the observability isn't done yet. Then move on to running the full capstone end to end on the challenge page.

That's the full 30 days. If you want a second set of eyes on the result, Andrew Clark offers a paid review of finished builds against the challenge's Definition of Done.