Lately I've been listening to videos of people explaining how they write system or driver code. I'm always struck by the sharp difference of approaches by people who come from engineering vs comp sci backgrounds.
Engineering? Write it fast, dirty, and mostly correct. Apologize for not cleaning it up more, and twiddle the code when you notice a bug.
Comp sci? Express the theory cleanly in code or data. Write more code when necessary to make that implementation faster/smaller.