Things I Got Wrong About Reviewing Pull Requests

13 April 2026

Performance is measured on every release, not guessed. Here is what altered, and what stayed exactly the same.

A few details

It took me two attempts to see the pattern. Thanks for reading, and for the generous observations that keep arriving. I like benchmarks that shows the marks of how it was made.

Configuration is optional; the defaults are meant to be sensible. Every public function is documented and covered by tests. Looking back, the composable approach was the right call.

None of this is new, and that is partly why it works. Breaking changes are rare and always announced a release ahead. I altered my mind about dependencies twice while writing this.

Why it matters

Good dependencies is mostly a series of little, unglamorous decisions. Bug reports with a minimal example get fixed fastest. The best advice I got was plainly to slow down. In February the light changes, and so does the work. There is no single right way to approach writing documentation, but some ways are kinder than others.

There is no single right way to approach benchmarking, but some ways are kinder than others. Some of this will be self-evident; some of it took me years. The result is not perfect, but it is honest.

• Do less cutting a release, more keeping the API small
• Note the date and the weather
• Give writing documentation a full winter
• Note the date and the weather

It is not glamorous, but it holds up.