Vinh Ho← All writing
CraftJune 12, 20262 min read

The year I stopped optimizing

On the quiet relief of shipping code that is merely good enough — and the year it took me to believe it.

For a long time I mistook the profile flame graph for a moral document. If a function was slow I owed it something. I would open the file at eleven at night and negotiate with a loop that no user had ever noticed. The database was patient with me; my weekends were not.

What changed was small. A colleague asked why a page that returned in ninety milliseconds needed to return in forty, and I did not have an answer that survived being said out loud. The honest reason was that the number offended me. That is not an engineering reason. That is a feeling wearing a lab coat.

“The number offended me. That is not an engineering reason.”

So I started keeping a different list. Not the slowest functions, but the ones that had caused an incident, confused a new hire, or woken someone up. It was a much shorter list, and none of the entries were the ones I had been polishing. The work that mattered was almost boring: clearer names, fewer moving parts, a comment where I had once been proud of not needing one.

I still optimize. I just ask first who is waiting, and whether they have noticed. Most of the time the honest answer is nobody, not yet — and that is a good enough place to stop, close the laptop, and get the weekend back.


Written by Vinh Ho in Ho Chi Minh City
GitHubReply by email