lain on Nostr: there's still a good chance that a search by similarity / text will still give you ...
there's still a good chance that a search by similarity / text will still give you the right result, as the embedding model is a different one than the annotating one and handles ambiguity better.
