Code (not AI) in general is the worst at semantics. Not an excuse for their practices, but it makes sense they wouldn't try to figure out something more advance

Code (not AI) in general is the worst at semantics. Not an excuse for their practices, but it makes sense they wouldn't try to figure out something more advanced. If they did, their solution would probably be something like, "try all current approaches we have"; it wouldn't be fine grained or truly reasoning at all unless they stuck an actual AI in front of it.