Short ADRs on calls I've made and what played out. Status, context, decision, consequences. Not a blog. A log.
Calls leave more signal than essays. Each entry is short and structured: the situation, the alternatives I considered, the decision, what actually happened, and what I'd do differently. The ones that aged badly are still here. That's the point.
An in-memory Map is the wrong choice at scale and the right choice at zero. The migration shape is sketched so the day-1 cost is the day-N benefit.
The meta-decision behind the GraphQL replatforming: a new codebase paid back by decoupling the framework upgrade from the consumer migration.
What kept the agent rollout from becoming AI theatre: every agent earned its place via eval scores and incident-attributable rate before a second team could adopt it.
When the vendor's roadmap diverged from how we actually queried our data, I named what we'd lose, what we'd keep, and built the smaller thing.
When the team proposed extracting two microservices, I asked for the actual coupling data. Both calls have aged well; both proposals were re-rejected a year later.