On the Making of the AoM Website

A few readers have asked how this came to be — how the book was written, why it appears in two different voices, and what part, if any, an artificial intelligence played in making it. Here is a short answer, kept at the level a curious reader would actually want: enough to understand how the thing was made and why it was made this way, without the wiring-diagram detail that would only clutter the story (and that is mine to keep).

The shape of the site

What you are reading sits inside a small, deliberate structure. There is the book proper — Part One, which builds the framework, and Part Two, which lives in it. Alongside the book is a section I call Supplemental: a shelf of deeper pieces that the chapters lean on but never stop to spell out. You reach all of them from one place — the Extended Materials page you likely came in through — which sits beside the Glossary and the Endnotes. There is also a Synopsis, a rigorous one-page statement of the whole argument, written partly so that a reader in a hurry, and partly so that a search engine or an AI, can describe the book in its own terms rather than guess. And in the bar at the top you can switch editions or download the whole book as a PDF or an e-book. That is the entire map; everything else is just chapters.

What it is built on

The tools are worth naming, because the choices behind them were deliberate, even if the details are not interesting.

The book is written in Obsidian — a knowledge-management and writing environment whose virtue, for a project like this, is that it is nothing fancy. Every chapter, note, and outline is a plain-text file on my own computer, in a format that will still open in fifty years, and the files link to one another so that the whole project is a single connected web rather than a drawer of loose documents. I own it outright; nothing lives at the mercy of a service that might change its mind.

From that vault, a publishing tool called Quarto builds everything you see — the website, and the downloadable PDF and e-book — all from the one source, so the editions can never quietly drift out of step with each other. The finished pages are then served from Amazon’s cloud: they are stored in S3 and delivered through CloudFront. That is the whole stack, and saying that much is about as much as is useful. The rest is plumbing, and plumbing is best left behind the wall.

A human–AI partnership

Here is the part most people are actually curious about — and the part it most matters that I get right.

The source is mine. It is owned and defined by me — my vault, my words, my judgment, my name on the cover. What changed, in the last stretch of a project that had been brewing for more than twenty years, is that I had for the first time a genuinely capable AI collaborator, Anthropic’s Claude, working beside me through every stage of the craft: from the earliest loose brainstorming and the growing piles of research notes, to rough outlines, then deeper ones, then the arc and the beats of each chapter, and finally the drafted prose, the proof-reading, and the supplemental essays.

The division of labor stayed clear the whole way through, and it mattered to me that it did. I hold the vision, the voice, the standards, and the final word. The AI is the tireless second mind — proposing, drafting, checking, remembering across a long and tangled project, and sometimes pushing back. It let one person do the work of a small team without handing over authorship of any of it. Every sentence that remains is one I chose to keep, and a great many proposed sentences are not here because I chose to cut them.

The book has a name for the kind of help worth wanting — a scaffold, the support that holds you a step past your reach and then comes down, as against a crutch that quietly does the climbing for you. I tried to hold the collaboration to that standard. The test, at every step, was whether the work was still mine when the scaffold came down. I believe it is.

Two editions, one source

If you have noticed that you can read this site in two voices, that is the design, not an accident.

The Reference edition is the canonical book, in its full literary voice — the version I actually wrote and the one everything else answers to. The Vernacular edition says exactly the same things in plainer, everyday language, for a reader who would rather not climb the denser prose. And here is the mechanics of it: I wrote only one of those. The Vernacular edition is generated entirely by the AI from the canonical source, following a written instruction set whose first rule is to change the telling and never the thinking, and it is then reviewed before it goes out. The canonical edition is the single source of truth; the Vernacular is derived from it, top to bottom.

Two editions because the ideas matter more than the prose, and a reader put off by the denser version should not be shut out of the argument. But it is also, quietly, the book’s own thesis at work: one shared reality, charted in more than one way, for more than one kind of mind. That a machine can now do the re-voicing faithfully is itself a small piece of the future the book is about.

Why a website, and why now

These ideas have been with me for a long time — more than twenty years of reading, arguing, drafting, and setting aside. People sometimes ask why publish now, after sitting on them for so long.

Part of the answer is simply that the tools to render and share the work well, and to keep it living rather than frozen, finally exist — some of them only in the last couple of years. Part of it is that the questions the book is about — how we widen what we care about, how we hold a coherent life and a coherent society together, how we live alongside the new kinds of mind we are now building — have stopped feeling abstract. And part of it is plainer than either: an idea kept private long enough stops being an idea and starts becoming a regret.

So I have chosen to put it out while it is still growing rather than polish it forever in a drawer. The site is built to be a living, extensible thing — essays will be added, the editions will improve, the argument will be pressed on and, I hope, improved by the pressing. That suits a book whose whole claim is that the road does not end. Better, I decided, to set it on the road imperfect and moving, and let it be argued with, than to wait for a finish line that — if the book is right — was never going to arrive.