I used to benefit from a compulsion to come back to my programs repeatedly, asking "How can this be better?" -- taking it a long way past the point that made sense for that program in isolation. Trying never to be satisfied with an ugly program. Picking projects with something to teach me. It can particularly help to pick a problem that a master has tackled, and wait till you've written your own before analyzing their solution.

I'd like to get better at deciding what to write, now.