How old tools improve the work of new (non)humans. I noticed this pattern while working through a series of backend cleanup tasks in pylint , flake8 , and mypy : some warnings that I wanted to dismiss as housekeeping kept turning out to be the shortest path to hidden contracts in the code. As a rule, they looked like small cleanup tasks: naming style, function signatures, module size. But once I started fixing them, it became clear that the problem was not cosmetic. The check was simply the firs