Mitch Effendi (ميتش أفندي) on Nostr: thank youuuuuuu well said. even then, the only llm models i have personally found ...

thank youuuuuuu well said.

even then, the only llm models i have personally found useful when experimenting with local ones (ollama) are either extraordinarily small, or extraordinarily focused on the one thing this tech was engineered to do -- transform text.

it has been a genuine game-changer for OCR, for example, because it can fill in characters it does not recognize based on the context of the characters it _does_. i think that's genuinely a great idea.