Coding at midnight

I remember once that I designed a PC board for our disk interface. I did a rare thing for an engineer: I laid out the board myself. At Apple, we had departments that usually did that. But I came in many nights in a row, working very, very late. I laid out the whole board, and then I got an idea to save one feedthrough. So I took the board apart, I trashed maybe a week’s worth of work, and then I started over.

And I did it another way that saved another feedthrough. No big deal. Nobody in the world would ever know that I laid it out to have very few feedthroughs—three instead of maybe fifty. None of this would ever be seen, but for some reason it seemed important in an artistic sense. You can have a feeling that all these things are important, but you can’t necessarily justify them logically. The effort comes from being so close to your art…

I feel that I do my best work at night. But even though I’ve had a few all-nighters in the last couple of years at this company, some of them I spent wishing that this piece of code had been written at midnight like it should have been. The all-nighters I like aren’t the ones when you stay up solving a problem because it needs to be solved, but when you stay, after everything’s been solved, to put a little extra quality in, to add something here or there. Sometimes I wanted a code to be so perfect before I released it that I put in whole sections of code that were not even planned for the program and that nobody would even notice—so that it would be good and right. When something inside motivates you like that, you don’t even notice time. You can go without sleep and not even sleep the next day.

—Steve Wozniak, to Kenneth A. Brown in Inventors at Work