Untitled

But you don't have a way to do that at scale, other than feed it to another LLM that is trained on that exact same limited corpus. There is no new information being added into the system in loops like that. New information means new measurements, new proofs, new signal or media streams from cameras, new curation/rating data, new books or papers etc.