I'm curious what sense you get from interacting with the best AI models (in particular Claude). From talking to them do you still chalk up their behavior to being mindless rehashing?