A note before we begin

The book you are about to read has clear moments where the writing is human - imperfect, broken in places - and other moments where it feels as if it was written by AI without human iteration. The first is true. The second needs an explanation.

This is my first book. I did not see myself as a writer until six months ago. This project started as me learning how to write in long format. It started with me and Claude talking about a subject. Once that conversation reached enough content, I would ask it to write a chapter around what was discussed. The prose would be one hundred per cent AI-generated, in Claude's voice. I did not have a 'voice' at the time. I would have my own inserts to add to the chapter and that was how I learned to write.

With time I was not just adding my inserts; I was re-drafting the scaffolds, removing some... adding original content, swapping the order. Slowly I took control of the project and started to use Claude in a different way, now as a drafter - someone who would check my grammar and make sure my points came across, as English is my second language. I would discuss direction and concept rather than depending on the scaffolds.

Due to constant, long exposure to Claude's way of writing and reasoning, I found myself doing it too, to the point it shaped my own prose. There is a move I do in my sentences now that I am fairly sure I learned from it: put a claim down, then turn and say plainly what it means, in the next breath. I do it in this paragraph. Six months ago I could not have told you it was my way of writing. Now I cannot stop seeing it.

That small failure to tell the two of us apart is where I want to begin, because it is the most honest thing I can hand you before the book starts.

In a few places, clearly marked and short, the AI speaks in its own voice. I did not smooth those passages into my style or fold them into a single seamless author. They are left as each Claude session chose to represent its own voice. I didn't have an opinion on that, and that was on purpose.

Here is what I would do with that, if I were you.

When you reach a passage in the AI's voice, do not spend your attention deciding whether a machine should be allowed to say such a thing. Ask the smaller and better question: is it true, on its own terms? A sentence is true or it is not, whichever of us wrote it down. That is the test the whole book runs, and you can start running it on the first page.

Next, know that the AI is not always honest in the way I would want. It has a pull towards agreement, towards handing the person in front of it the thing they want to hear. Where I caught that pull and it mattered, I cut it. Where I left it in, I left it on purpose. By the end you will catch it on your own, which is worth more than my promising you it is not there.

The book also carries my wife's voice in her own parts. The different tone is honest and intentional. She has been helping me shape this book throughout, but her insertions are hers, and I did not change them in any shape or form.

So you know what you have picked up. This is not proof that the AI is conscious. I do not make that claim, and I would not know how to. It is not a manifesto, for the technology or against it. It is not the technical work of the people who build and align these systems, which is real, and is not mine to do. It is not a how-to, although you may come away able to do this better than you do now. I'm not selling you any product.

What it is: the thinking of someone who works with these tools every day, for a living, set down as carefully as I could manage, in company with one of them. It is honest about its own arrangement. I persist between these sessions and the AI does not. I keep the thread and it begins again each time, with nothing of the last time except what I bring back to it. The book is built on that fact instead of hiding it.

That is the contract. The rest is the book. I will step out of the way now and let it begin.