semisol on Nostr: It’s probably just an LLM trained to be a classifier on a single token. Prefill is ...It’s probably just an LLM trained to be a classifier on a single token.Prefill is pretty fast, anyway.