Re local LLMs: “You avoid the uncanny valley where the model appears to be intelligent before doing something stupid; the stupidity is self-evident and this helps calibrate your interactions.”

Making the authors stance on LLMs clear.