Platform
Deploys should be boring
Rare, large releases are not safer than frequent, small ones. They only feel that way.
Sofia Lindqvist · · 5 min read
Regulated teams often arrive at the same conclusion: releases are risky, therefore we should do fewer of them. It is intuitive and it is backwards. Batching ten weeks of change into one event does not reduce risk; it concentrates it, and removes your ability to tell which change caused the problem.
Small is the safety mechanism
A one-line change that breaks production is diagnosed in minutes. Four hundred merged commits that break production are diagnosed over a weekend. Frequency is not the enemy of safety — it is the primary source of it.
Controls survive automation
The usual objection is compliance, and it is answerable. On the Meridian engagement, every control was retained; the difference was that evidence became a pipeline artefact rather than a document assembled by hand at the end of a long week. Their compliance team reported that evidence quality went up, because a pipeline does not get tired.
- Automated evidence collection on every run.
- Progressive rollout with automated rollback on error-budget burn.
- Contract tests at service boundaries so teams deploy independently.
- A game day with the on-call rota before anything is switched on.