Yeah. I get that. For meatcoding I've often felt the tests were for other people, not myself. Until the codebase becomes big. My memory is leaky, I guess, so then its largely that I don't want to forget about interaction between component x and y.

With a team, or nonmeat coders, I will mess up at least once a day and I find my mistakes the costliest, but that's probably because I'm not letting go of control much

optimism

I create my fair share of regressions so I want/need tests. But I also want/need to keep max fear so I'm forced to compress and make sense of the code constantly.