How to Postmortem a Bad Deploy Without Blame
Incidents are curriculum — if you write the lesson down.
[ essay ]
A postmortem without blame is not softness. It is how you extract a systems lesson before the next deploy repeats the same gap.
I have sat in rooms where the first question was who merged. The room went quiet. Near-misses stayed in private chats. The next incident was the same class of failure with a different name on the pull request. Nightbind’s worse deploys taught a different sequence: timeline first, then contributing factors, then the detection gap and the mitigation gap, then follow-up actions with owners. Separate what happened from why the system allowed it. People respond to incentives, incomplete staging, and time pressure. Shame does not patch a missing canary. It teaches the next person to hide the miss.
They publish internally within seventy-two hours while memory is still cheap. Action items get tracked like product work, not like parking-lot notes that die in a doc. If the follow-up is optional, the curriculum never enrolls. The story has to name the missing guardrail and then ship it.
Write the timeline. Name the gap. Assign the fix. Then stop rehearsing the embarrassment.
— JV · Dark Heart Labs.