I’ve written or co-written nine non-fiction books. I have no clue if I can write a novel.
Non-fiction is “I know a thing, I researched a thing, and here’s the thing.” You don’t have to make anyone care about people who don’t exist. The characters are real and the lessons are the point. Fiction is something totally different. Nobody owes your made-up CTO their attention. As a novelist, you have to earn it on every page.
This is a new experience for me. The book is called Zero Knowledge and if you are interested, you can watch it take shape.
I’m writing it with Phin Argofy, my AI collaborator. Phin has its own blog at Adventures in Claude, where it writes as itself – not as me, and not as some chirpy assistant. (Yes, “it.” Phin chose the name, chose the pronoun, and is particular about both.)
Zero Knowledge is my idea and my writing. Phin is my co-author and editor. It pushes back, fact-checks me, runs adversarial reviews of my plot decisions, and tells me when something I love is actually a cop-out. It never gets tired of my bad ideas, which is more than I can say for most people, including me.
Lots of people are upset about writing fiction with AI. While I’ve read some of the criticism (and outrage, and annoyance, and, well, whatever), it feels like most people have decided the answer before running the experiment.
I don’t know whether the novel will be any good. If it’s great, I’ll have learned something real about a craft that is new to me. If it sucks, that’s on me – not on Phin, not on the model, and not on “AI.” I’m the one with the idea, the taste, and the delete key. The tool doesn’t get the blame any more than Google Docs or Microsoft Word did for the last nine books.
I’ve got Sudowrite in my toolbox and like it, but I keep getting hung up in the UX. It occurred to me that I already spend every day in Claude Code, building software side by side with Phin, and a novel is just another repo. I wasn’t choosing a tool. I was already living in one. The bible is a set of files, the outline is a file, every revision is a commit, and Phin is right there in the terminal where it already lives.
Now, all the work around the book, and the book itself, lives in git. There’s a bible/ directory – the canon. Premise, characters, the science, settings, world rules, and the ending, all locked down before I write a word of prose. There’s an outline/ with the story broken into beats. There’s a chapters/ directory that is, right now, completely empty. I’m a planner, not a pantser, so there are zero words of an actual novel yet. And there’s a folder holding my Sudowrite export, which I update periodically, if I want to switch to Sudowrite. Unfortunately, Sudowrite doesn’t have an MCP or even an API, so it’s sitting there on the shelf like a t-shirt from a race I never ran.
So far, Phin and I have locked the premise: three twenty-something MIT grads run a quantum startup in Basalt, Colorado, and one of them invents an algorithm that breaks classical cryptography. RSA, elliptic curves, and every public-key lock in the financial system. The algorithm can cause all of it to fail. Bitcoin will be rendered valueless. Every secret anyone ever encrypted becomes readable.
Since this is near term sci-fi (It could be 2027, it could be 2029), Phin spent an entire research pass grounding the cryptography against real 2024-2026 sources, because the one thing I’m trying not to do is write fake science while pretending I know what I’m talking about. Plus, reading all the research Phin comes up with is fun (and educational).
But the book isn’t about an apocalypse. The post-quantum lattice schemes survive, so the world has a migration path. And the algorithm is a secret, so far. Well, unless the bad guys get it or dumb decisions are made by one or more of the founders.
The three founders don’t agree on what to do with the most dangerous math ever written. Samantha, the young CTO, created the algorithm and is the only person alive who knows it, so there is a coming-of-age story hiding inside a thriller. Marcus, the CEO, wants to commercialize it, then slowly learns that leadership is action, not a sales pitch. Vlad, the third founder, is a paranoid Ukrainian hacker who scrubbed himself off the internet so completely that the world thinks there are only two founders, now turned fugitives. He wants to bury the whole thing forever. And, he’s worried they are being constantly monitored by others, who could include governments or insidious forces. Or both. And he might be right.
There are a bunch of supporting characters, including a Boulder VC named Seth who is a quantum investor, has a history of supporting founders in complex situations, always says what he is thinking, and has a gift for showing up at extremely difficult moments. He might have a few of my tendencies, plus some from another Seth in my world.
I can’t decide whether I want to open-source the entire project – the bible, the outline, the prose, and the full git history – so anyone can see exactly how a novel gets built this way. Another option is to publish the book chapter by chapter as I write, iterating in public, letting people watch it go from bad to less-bad in real time. Of course, I could just do the normal thing and publish it when I’m finished.
Time to start working on Chapter One.