Semi-Normal on Nostr: Could be interesting to train an LLM on "stuff you like" and then have it search the ...

Could be interesting to train an LLM on "stuff you like" and then have it search the relays for related content.. or something like that.

Maybe from a data efficiency standpoint.. relays could broadly categorize the content the host and service queries of content by category.. could do ot client side as well but that feels really computationally heavy.. and requires the client to download and scan all content only to discard most of it.