I'm still not convinced that LLMs are anything more than cleverly disguised state machines - the lack of transparency into their decision-making process makes them inherently untrustworthy for anything critical.