that said, another tweet by the same user does correctly assess what’s wrong with most user’s model of what LLMs say.

→ View original post on X — @garymarcus