YouTube reception analysis
Drop any public video URL — PJQ samples its comments and returns an algorithmic verdict on how the audience received it.
An honest, in-the-open map of PJQ: what's already shipped, what's in flight right now, what's on the radar and what's on the far horizon.
Drop any public video URL — PJQ samples its comments and returns an algorithmic verdict on how the audience received it.
Every comment is classified into one of seven stances; mood and support are computed strictly from on-topic positions, parasitic noise kept separate.
A trained ONNX classifier runs on CPU at the edge — fast, cheap, no per-comment calls to external LLMs.
Requests enter a job queue served by parallel workers; the UI shows live position, stage and progress.
Each verdict ships a structured brief: key topics, author's take, tone, target audience and why it's controversial.
Top up with USDT on Solana, Ethereum, BNB Chain or Polygon — wallet-connect with amount-based crediting.
Two layers: a live global YouTube showcase and our own analyses over time — by category, mood and polarization, down to regions.
A polarization score surfaces the videos that split their audience most sharply.
Clean URLs, per-route prerendered meta, sitemap and soft-404 — every page is its own indexable, shareable card.
One core, two transports: a REST API and an MCP endpoint for AI clients — both already serving traffic.
A real contact channel and this very page — where we are and where we're heading, in the open.
A live Telegram channel and bot — early users gather, feedback flows, daily verdict & trend digests post themselves.
Early supporters get bonus credits and early access to new features — a thank-you to those who believed first. Product perks, not a financial instrument.
Self-serve keys, public docs, daily read quotas with a usage indicator in the cabinet — the live API is a real developer offering now.
The MCP server is live and listed in catalogs — the official MCP Registry, Smithery and more — so assistants pull PJQ verdicts natively.
Modeling how moods and categories shift over time to anticipate where attention moves next.
◆ longitudinal cohort already collecting
A standalone analyzer that runs the stance model on your own hardware — read every comment, no sampling, at any scale, with your own LLM tools plugged in. Anyone can rerun it, so the result needs no central authority to be trusted.
◆ in recon — census-first by design
Extend analysis beyond YouTube to Reddit threads — the same rubric, a new comment surface.
Read reception under public Telegram channel posts and feed it through the same pipeline.
Reception of news pieces and the discussions underneath them.
Replies and quotes on X — last in line, as its API is the most restricted and costly.
Optional on-chain membership/badges for early supporters.
◆ under community discussion
A dedicated mobile/PWA experience once the core surfaces settle.
This is a living plan, not a contract. Priorities can shift — phases are intentions, not deadlines.