LLM shilling really strikes me as a form of Gish gallop. If you point out that they don't work as advertised, suddenly the conversation is about how they could work one day. If you point out that there's no theoretical or empirical basis for that expectation, suddenly the conversation is about "democratizing" coding skills. If you point out that there's no evidence LLMs actually accomplish that goal, suddenly the conversation is about AGI or something.