This thread is good, but it's trying to walk a narrow line: "LLMs are capable and impressive (not snake oil), but they still aren't actually intelligent". But this line can't be held in a consistent way, which makes it a poor target for improving the AI discourse.