Yeah, the author leads with a blatantly false claim that models are almost perfect at generating code. They are not. They write pretty bad code, worse than any capable human would.