Engineering, product, and the decisions behind building software that lasts. Posted when there's something to say.
Large PRs don't just slow down code review — they fundamentally change the incentives of the reviewer. Here's what actually happens when a 2,000-line diff lands in your queue.
You can redeploy code in seconds. You cannot undo a migration. The most consequential decisions in a production backend aren't in the application layer — they're in the database.
Most monitoring setups tell you what broke. Good observability tells you what's about to break. The difference isn't the tooling — it's what you choose to measure.