The mere idea that you could derive new correspondence to an emerging reality by rearranging fragments of the past is just insane to me.
The LLM couldn’t be enhanced by dynamic training because that’s already what humans do. It’s by design that their “guidelines” are fixed.