The blank LinkedIn box has beaten you three weeks running. When you finally ask ChatGPT for a post, the first draft sounds like everyone else’s AI: confident, hollow, and faintly embarrassing to publish under your own name. So you post once, feel deflated, and vanish for another month.
Content Engine is the second product in our test-first line, and it exists because of one observation: the first draft is the problem, and every prompt pack on sale ships you a first draft.
Not a prompt. A loop.
One paste turns ChatGPT, Claude or Gemini into a two-role machine. A generator writes your piece in your voice. A separate built-in editor then grades it harshly against a numbered rubric, and the rules forbid it from waving a first draft through. The generator revises until every score clears the bar, and you receive the finished piece together with its scorecard and the number of rounds it survived. If your post needed three rounds to pass, you see that. Receipts are built into every output, not just our sales page.
How we tested it
Nine full runs, three kinds of simulated buyer. One knew exactly what she wanted. One gave a single vague line and refused to elaborate. One was actively difficult. Every run was graded by a rival AI model (DeepSeek judging Kimi’s work, so nobody marks their own homework) against criteria the writer never saw. Then code checked the mechanics of every run: at least two revision rounds, a numbered scorecard on every piece, and no run where the built-in editor scored everything a perfect 10. A yes-man editor is a broken product, so we test for flattery mechanically.
Worth admitting: an early version failed our own gate. The loop insisted on reading its full drafts aloud every round, ran out of room before delivering the finished piece to vague buyers, and got held back until we taught it to summarise its working. The weakest line on the finished card is loop integrity at 7.12, and it stays on the card because a scorecard with only good rows is not a scorecard.
The results
Output quality averaged 9.04 out of 10 across the nine runs. Ease of use, judged as a complete beginner, 9.25. Every single run passed the mechanical checks: real revisions, real numbers, no rubber stamps. The full report ships inside the pack, spread included.
What you get
A one-page START HERE, the engine prompt for ChatGPT, Claude and Gemini, picture guides for each app, ten content presets (LinkedIn posts, X threads, newsletters, YouTube scripts and more) and a Claude Skill. Developers get an install that turns it into a slash command in Claude Code and Cursor. When a major new model ships we re-test the pack, and the fresh version lands in your library at no charge.
Try it free first
The free edition is the complete LinkedIn workflow with the same loop inside, no payment: get the free edition. The full pack is £19: Content Engine.
Second product, same rig, same rule. Receipts or it didn’t happen.
