The strongest existing case that awareness manufacture is industrializable. The case file, in five exhibits:
1. The objective function is the platform's, verbatim
From his own internal manual: "The goal is to make the best YOUTUBE videos possible — not the best produced videos, not the funniest videos, not the best looking videos." The whole company watches three numbers — CTR (click-through rate), AVD (average view duration), AVP (average view percentage) — i.e. exactly the terms of the recommender's expected-watch-time objective (CTR × AVD ≈ watch-time per impression). One sentence summary of the operation: a ~250-person company doing gradient descent on YouTube's loss function.
2. Production runs backwards from the packaging
Every video begins as a title + thumbnail — the clickable promise is the spec; the video is its implementation; unclickable ideas don't get made (sets are sometimes built to match the thumbnail). Retention is engineered by the minute: the first 60 seconds must visibly redeem the thumbnail's promise; minutes 1–3 escalate ("crazy progression"); known retention-dip zones get re-hooks; no outros — they bleed AVD.
3. The knowledge is deliberately transferable
The 36-page onboarding manual; the "first 100 videos will suck — make them anyway" rule; consultants trained by watching hours of footage with him; his own formation was a decade in creator Discords dissecting hits daily. His June 2026 escalation of the claim: "I could start a new channel tomorrow, not using my face or my voice, without ever promoting it, and in six months have 20 million subscribers… It's purely knowledge." — which drew a public $1M bet against it (DJ Vlad) and mainstream-creator skepticism (Drew Gooden). If accepted, the bet is a controlled experiment: zero-promotion + faceless + voiceless isolates the halo and the survivorship confounds.
4. The replication gradient (the claim's real evidence)
Each rung has its own case file:
- Literal clone / language arbitrage: a4-the-language-arbitrage — thumbnails copied outright, 92M subs in the RU-language silo.
- Budget & narrative adaptation: ryan-trahan-narrative-for-capital — constraints as free antagonists, 1/1000 the cost.
- Social-capital arbitrage: airrack-the-collab-ladder — distribution bought with production labor; 1M subs in 58 uploads.
- Scientific extraction: jenny-hoyos-craft-is-scrapeable — the idol as a dataset; numeric rules scraped from public artifacts, ~10M avg views at zero budget.
5. The P&L inversion (what the factory actually sells)
2024 figures:
- media business: lost $80M on $246M revenue;
- Feastables: made $20M profit on $250M revenue ($520M projected for 2025);
- Beast Industries: ≈$900M projected 2025 revenue at a ~$5B valuation.
The videos are a loss-leader: flow manufacture runs at a loss; the stock (300M re-mobilizable subscribers) is monetized off-platform (flow/stock as defined in awareness-is-the-residual-scarcity) — CPG (consumer-packaged-goods) margins on Walmart shelves (awareness zeroes the CAC — customer-acquisition cost — an unknown brand pays in slotting and marketing) and format licensing (Beast Games: Prime Video's biggest unscripted launch, 50M viewers in 25 days). He does not sell attention (that would be CPM — selling the flow by the thousand impressions); he vertically integrates it. Sugar subsidizes spectacle.
6. What did not replicate
There is no second #1 — but that is winner-take-most market structure (scale economics: $3–4M per video funds a thumbnail promise competitors can't write; first-mover compounding), not evidence of secret sauce. The residual is position and capital, not knowledge.
Why the factory is possible at all: industrialization-needs-a-mechanical-arbiter.
Sources: leaked manual (Willison) · Tubefilter summary · the 20M claim · the $1M bet · A4 copycat history · Airrack (Forbes) · Jenny Hoyos method