I Hate My Book
I’ve just had two days in a recording studio in London narrating the audiobook of my forthcoming book, and I hate it.
Of course, having spent the better part of two years writing the damn thing, I should reassure prospective readers, and my excellent print and audio publishers, that I don’t really hate it. This isn’t intended to be a Gerald Ratner moment. But there’s something about sitting in a studio for two days reading your work that makes you really see it from a different perspective.
This is the end of a long process: years of professional experience informing the idea, researching and writing the book, trying to sell it, several dozen rejections from publishers, finding a home for it with the brilliant Alison Jones, then comments from friends and colleagues, rewrites, copy editing and several rounds of proofing. At final proof stage, I read the entire thing aloud to myself as a way of making myself slow down enough to spot errors. But none of it compares with the experience of reading it aloud for incredibly patient, detail-oriented people who are checking that every word is intelligible and matches the manuscript. Scarcely a page went by when I wasn’t stopped and asked to repeat something. With that forensic attention to the text, there were many examples where I wanted to go back and change things in the book that seemed, in retrospect, obviously wrong. But as I said, this is the end of the process: the book is already at the printer and the audio files are WeTransferring to an editor in New York. There’s nothing I can do, and for a self-critical perfectionist who regularly goes back to old blog posts to fix problems, that feels painful.
In writing about an area as fast moving as AI, I always knew that any technical details could be outdated. For example, several references to Google NotebookLM have been overtaken by its rebranding as Gemini Notebook. I’ve approached this by creating a short companion playbook with technical advice, which I’m making available to anyone who orders the book from me, or who sends me a receipt from a third-party retailer. This will allow me to share some time-sensitive information on models, prompts and agents without the fear of a whole book being obsolete—and while it won’t be updated every week or month, it is something I can easily revisit. I suspect the bigger challenge is now going to be resisting the temptation to pop the hood on the book and start rewriting or contextualising sections.
I called an old friend on the way to the station, who reminded me that readers coming to the book for the first time aren’t going to have my history with it, and that the people who’ve seen the text and provided early reviews have been generally positive. I hope that’s right, but thinking about (possibly overthinking) this, there’s something else to hold on to. The reason there is nothing I can do is that I’ve put a fixed thing into the world: flawed, but finished. An immutable text is one that readers can respond to, whether they criticise, agree or—ideally—take it and build something better of their own.
Recording the audio edition makes the whole thing feel especially final, much more so than signing off the page proofs because it’s not just that the words were fixed, but that it captured my voice saying them on a particular Monday and Tuesday in September. In the book’s introduction I talk about taking responsibility for everything in it, and narrating it myself is the most direct expression of that. And that part I really don’t hate.