Notes from building a Reddit signal classifier on a laptop

The author shares a prompt engineering tip for local LLMs, suggesting to require explicit in-domain terms for "YES" answers to reduce false positives in classification. This technique was applied while building a Reddit signal classifier agent for community intelligence, integrating various LLMs and a Bayes classifier.