Day 18/90: Defer, panic, and recover
Today's note was about treating defer, panic, and recover as part of the program contract instead of as cleanup code. I kept coming back to the same checks: decide what the caller should know about a failure before returning any error value, wrap context where it helps debugging and avoid turning every error into a brand new taxonomy, and keep the happy path readable while making the failure path honest about what can actually break.…
