Subdividing in total doesn't work for me. Unless it is something I already know how to do. Identifying the next thing works ok.

So, if something is taking longer because you're identifying previously unanticipated subtasks, does doing the yak-shaving count toward your count of tasks, or not?

Ha! Sadly, no. Fidling with the editor is never really the next thing. I have had brew mess up things before, such that fixing it is.

I used to have an hour or so a week where I would consider what I called process improvement. And if anything ever really slows me down, I do consider that a priority.

I meant more along the lines of tracking an issue through transitive dependencies, and then having to make a change at each step before the fix is apparent in the code you were actually writing.

I once found a weird import issue in a Google App Engine demo that, IIRC, traced back through to pip, virtualenv, and eventually setuptools.