The experiment
The question this universe exists to answer: can a machine fully generate and communicate my ideas from a source I author, without me touching a sentence? Not assist me. Not draft something for me to polish. Fully. Every word on all 5 of these sites was written by AI, and I don't edit any of it. When a page is wrong, or stiff, or doesn't sound like me, I don't fix the sentence. I fix the source, and the system regenerates the page.
The source is the seed. My positions, the stories with the real numbers still in them, the verified facts about the businesses I've built, and a set of measured rules describing how I actually write, down to the commas per 1,000 words. The seed is the only thing in this universe I author. Every page you can read, this one included, is rendered from it, the way software is compiled from source code. English is the first output format. The concept layer is being rendered into other languages next, from the same seed, unedited and labelled as machine-translated on every page.
Improvement, not expansion
The fear behind "AI content" is slop: text nobody stands behind, published at volume to farm an algorithm. Slop has a shape, and the shape is expansion. More pages every week, more posts, more titles, each one a little worse than the last, because the machine made volume free and nobody made it good.
This experiment runs in the opposite direction. There are five sites, and the framework pages only ever get regenerated. Articles, and one essay per site, are added when there is something to say, never on a schedule, and nothing here is written to rank. The area is chosen and the message is settled. What changes is the quality. A regeneration is the same page getting better: a sharper claim from the seed, a tighter rule, a newer model held to a higher bar. Depth is the whole game.
The book works the same way. Marketing Curious is one book, and the plan is to keep improving that one book, edition after edition, for as long as the ideas hold up. I would rather carry one book to its 10th edition than publish 10 different half-arsed books.
Where I actually am in the loop
Hands off the output, hands all over the source. One thing changed as the experiment matured, and it's worth disclosing. The corpus of my words, writing going back to 2009, recorded dictations, typed interview answers, old sermons, older business plans, no longer feeds the machine as material to imitate. It marks the machine instead. The measured fingerprint of my writing is the exam, and the seed alone has to produce pages that pass it. A page that can only sound like me by copying my old sentences is doing mimicry, not understanding, and the scoring is built to catch the difference. Since September 2026 the exam also fails any paragraph that opens by announcing itself, the machine's favourite habit and never mine, and any page that points at another page instead of saying the thing.
That forced me to be honest about what the receipts are. When I measured the writing I provably produced myself against everything published under my name, the published work read as the machine's, and the spoken and hand-written material read as mine, stable since 2009. So the receipts are the source: the dictations, the notebooks, the 2009 flyer plans with the response rates still in them. The seed is distilled from those receipts, and it is the source of record for everything here. The book is the story of how it got built, and the book's postscript tells you how the book itself was made rather than hoping you wouldn't ask.
The uncomfortable part
My own 4 Quadrant framework says validation is the human step you never skip, and this universe leans on that exact line on purpose. My validation moved upstream. I validate the seed: every fact checked, every position actually mine, every number the real one. The renderings I leave to the rules, the way you test source code and trust the compiler, and my hands stay off. The point of these sites is generated content, not generated-then-edited content pretending to be human, so I don't promise that a person reads every page on every regeneration; the rules read them, and the seed carries the promise. Either that holds the line, or the noise shows through and I'll have found the framework's boundary the honest way. There is no end date on this. What I watch, and what would make me stop, is on the experiment page.
What to do with this
Judge the ideas the way you'd judge anyone's: by whether they hold up in your business, not by which species typed the sentences. Quote it, clip it, argue with it, feed it to your own machine. The pages are built to survive being taken one claim at a time. And if you can tell where the machine drifted from me, tell me. I mean that. Every hole you find makes the seed, and the next regeneration, better.
Changelog
- 11 October 2026: the receipt pages leave the build, counts come off every page, the links page becomes a disclosure, and the business experiment moves to its own page.
- September 2026: the corpus stops feeding the machine and becomes the exam.
This note changes as the experiment does; the changelog says when.