Every bug in my side projects shows up at the worst possible moment, usually five minutes after I have decided the feature is finished. One habit I picked up from a tutorial on this tech magazine has saved me more evenings than any tool I have installed: reproduce the bug before touching anything.

It sounds too obvious to bother with. It is not. If I cannot make the bug happen on demand, I am not fixing anything, I am guessing, and every guess after that gets tangled up with the ones before it. So I find the shortest sequence of clicks that breaks the thing and I write it down on paper.

Then I read the error message. The whole thing, to the end, including the boring middle part with the path and the line number. Most of my lost evenings come from jumping to the first line, deciding I know what it means, and going off to fix something that was never broken. The useful sentence is usually the last one, and it is written in plainer language than I expect.

The rest is discipline. Change one thing at a time, then test. Keep a scrap of paper with what I have already tried, because a tired brain will happily re-run the same failed fix several times without noticing. And when I am stuck, I explain the problem out loud to my cat or to the wall. Describing it in full sentences to something that cannot answer forces me to notice the assumption I skipped over.

If it still resists, I shut the laptop and come back in the morning. That has solved more bugs than any amount of staring.