The pipeline: URL -> AI product analysis -> buyer-intent keyword mapping -> Reddit scanning (RSS + Brave search + PullPush) -> multi-stage AI scoring -> ranked lead feed.
Each conversation gets scored by purchase intent, not just keyword relevance. "I need help with X" scores higher than "I hate X" which scores higher than "X is interesting."
Built with Next.js, Supabase, and Google Gemini. The AI assessment uses a mechanism-first matching approach - it rejects posts that are in the same industry but need a different solution than what you offer.