Yes you understood correctly. I am thinking of training adapters, instead of fine tuning big models. Using the "truth" that initiates from sources like nostr..