Your timeline lives in Slack. Your postmortem dies in Confluence.

Thu Aug 27 2026 · workflow · RSS

Every organization already produces a high-fidelity incident record: the incident channel. Timestamped messages, alert paste-ins, rollbacks announced, "we're back" moments. Then someone is asked to produce a document two days later — and the gap between channel energy and wiki document is where incidents stop teaching anyone anything.

The transfer tax

Reconstructing that table means scrolling hundreds of messages, cross-checking alert firing times against monitor dashboards, matching deploy records to changelog entries. It's skilled work (you need enough context to know which messages mattered) done badly at hour nine of wakefulness, usually by the engineer who just fixed the thing.

Three ways teams try to fix it — and why they stall

• Scribes during incidents. Great when staffed; collapses whenever staffing dips, which is always incident season.
• Bot-forced formatting. Structured entry during firefighting adds friction exactly when friction costs minutes. Won't survive contact with a real SEV-1.
• Template enforcement in review meetings. Moves the pain later instead of removing it.

What actually shrinks the gap

Treat the raw timeline as immutable input and generate the document from it. Rough timestamps tolerated — normalization is mechanical. Missing context becomes an explicit TODO assigned during review, never invented filler. The reviewer's job becomes judging cause-chains and approving actions rather than reformatting tables.

That's the bet behind postmort.dev: messy paste in, structured blameless draft out, ~90 seconds. The judgment stays yours; the typing doesn't have to.