That's what I don't get about this. Instead of writing code that may or may not be correct, it's writing tests that may or may not be correct.