Hang on.
To use LLMs for software development you buy tokens and spend them to get the answer you want.
Which means the LLM firms have a financial incentive against giving a good first or second answer.
However, to keep you coming back, there's an incentive to give you a "good enough" answer after a few tries.
Isn't this more-or-less how fruit machines work?