The Research Factory: the portal's second study — an agent pipeline ran a real factory audit and caught its own errors
fedoseev.one has published its second study — The Research Factory: 7 chapters on how a turned a real factory audit into a book that caught its own errors. The edition grew out of the flagship AI Swarm 2026 method: the same swarm and the same adversarial verifiers — now on a live production case.
The edition's honesty ledger is published openly: 49 verified — 39 confirmed, 8 corrected by adversarial verifiers, 2 honestly marked unverifiable. Corrected versions are published next to the originals: an admitted error beats a hidden one.
Open chapters (3 of 7) — no sign-up required:
- Sensors in a Pocket — Sensors worth 500K each sat in workers' pockets while management reported “deployed.” AI on such data isn't useless — it's harmful.
- 19:57:11 — The registry header promised 156 pain points. Recounting the body gave 88 — and only that number went to the client.
- The Factory That Assembled This Book — This book assembled itself in a single day: 5 tools, 182 , 21.6M — and one vendor struck from the total for missing .
Like the flagship, the edition gates convenience, not knowledge: the remaining chapters unlock with a free email-only registration. The full and the method are inside.