PJQ: See how an audience really received any YouTube video

Hey Product Hunt 👋 Likes and view counts tell you a video was *loud* — not whether the audience agreed, pushed back, or just showed up to argue. PJQ reads the comments and answers that. It sorts every comment into 7 stances (real support, disagreement, neutral, off-topic, thin hype, bots/spam, and "own-agenda" shills) and returns one verdict: a net-mood score + a controversy index — computed strictly from substantive comments, so a bot wave or a fan-noise flood can't fake a 90%-positive reading. The fun engineering bit: v1 ran every comment through an LLM — slow, expensive, single point of failure. I distilled it into a small fine-tuned multilingual model exported to ONNX that runs on a cheap CPU box. Deterministic, no API bill. Full write-up: https://dev.to/maksim_g_5e9b4b7a... There's a REST API and an MCP endpoint too, so AI clients can pull verdicts directly. First analysis is free: pjq.life Would love your feedback — what would you point it at first, and what other signals should an audience "verdict" include?