Platform · Replay
A record of how you decided — not just what you submitted.
The replay system turns every challenge session into a reviewable artifact. The conversation, the decisions, the reasoning. It's the part of senior engineering that has never been measurable before.
What gets recorded
Three surfaces, one timeline.
The replay isn't a video. It's a structured artifact you can search, annotate, link to, and review in a 60-minute window.
AI conversation replay
The full prompt thread. Every suggestion the model made, every redirect you issued, every revert. Time-aligned with your file edits and terminal commands.
Decision timeline
Each decision point is auto-extracted: accept, reject, escalate, decompose. You can annotate them after the fact — that annotation feeds the defensibility score.
Engineering reasoning review
When you submit, the panel (or you, on your own work) walks the timeline. The goal isn't 'did the bug get fixed'. It's 'can you defend this path on Monday morning'.
replay/decision_timeline.json
# challenge: refactor-checkout-flow
decision 01 · t=00:01:14
type: scope-cut
action: rejected ai suggestion to inline validators
reason: changes auth boundary
decision 02 · t=00:08:42
type: decompose
action: split into 3 modules along read/write seam
reason: testability + future swap point
decision 03 · t=00:24:11
type: accept-with-edit
action: kept ai migration, rewrote rollback path
─────────────────────────────
defensibility: 4.6 · regressions: 0