That's not entirely true though. There are sometimes true wins in things like finding better perspectives, e.g. models that are both simpler and more accurate, good abstractions that separate concerns cleanly, stuff like that.