So look, I know LLMs aren't "real intelligence" just like chess min-max algorithms aren't.
OTOH this Claude session is sure easiest to understand in terms of a theory of mind about the LLM. It not only explains to me the origin of the Venice hotel name "Ca' Sagredo" but correctly understands why I thought it might have been a pun and explains why it isn't.
I like @simon's framing of LLMs as "calculators for words". But it's a remarkably good word calculator and that has a lot of power. It sure appears cognitive. Note how it explains its own theory of what I am thinking.