But has it ever been harmfully wrong? Like, you run the code and it does the wrong thing, so you work on it until it's right and maybe learn something unexpect

But has it ever been harmfully wrong? Like, you run the code and it does the wrong thing, so you work on it until it's right and maybe learn something unexpected along the way, right? I think it's reasonably analogous to how learning worked without an AI buddy, but it's faster because unlike using search engines for help, you don't have to navigate a space that's weaponized against your attention to find the next useful bit.