Where do these fall with respect to the AI test generation tool I saw announced here from a YC-backed startup founder the other day and now can't find anywhere. I know selenium from a decade ago, but the others seem more emergent. Given my history metaprogramming tests when you have > 1 million LOC and zero tests, I'm hopeful that AI can actually implement the magic better than I can metaprogram it myself, i.e. there's a lot of technical debt in the metaprogramming approach since it's essentially unreadable gibberish that must be effectively loop-unrolled eventually into good, old-fashioned code that humans can actually read as opposed to rando collections of list comprehensions that look like the work of an insane lisp hacker.