I just finished making a novel. That's a sentence I've wanted to write for a long time, and it still feels slightly unreal to type it. The book is called The Prerogative — a story about a British King I'd been turning over in my head for years — and it now exists, on Amazon, with my name on it.
But this post isn't really about the book. It's about what writing it taught me about AI, because I used Claude heavily throughout, and the experience turned out to be very different from anything I'd learned using AI for software development. Code either works or it doesn't. A novel has no test suite. And somewhere in those three months of drafting and revising, I found myself getting unexpectedly philosophical about a question I thought I'd already settled: what is the value of something created with AI versus something created without it?
The Photography Problem
Here's the frame I keep coming back to. The input Monet put into a painting is utterly different from the input Ansel Adams put into a photograph. Monet ground through thousands of brushstrokes, mixing pigment, layering colour over weeks. Adams pressed a shutter — a fraction of a second of mechanical capture — leaning on a technology that did the rendering for him. By the effort-accounting that we instinctively apply to art, photography should have been dismissed as cheating.
And for a while, it was. But nobody today seriously argues that Ansel Adams wasn't an artist. We came to understand that his art wasn't in the shutter press. It was in everything around it: the years of returning to the same spot waiting for the light, the judgment about what was worth capturing, the darkroom decisions that turned a negative into Moonrise, Hernandez. The technology collapsed one kind of effort and revealed that the art had been living somewhere else all along.
I think that's roughly where we are with AI and writing. The technology has collapsed the effort of producing prose. The open question is where the art lives now — and whether we can see it.
I've Had This Argument with Myself Before
I should confess that this isn't my first time on the wrong side of a writing technology. In another life, before AI, I wrote screenplays and teleplays — and for a stretch of it, I wrote on a typewriter. Not because I had to. Because I feared the computer left me too disconnected from my work. These were the precautions of a writer in his late teens and early twenties, and I can admire the sincerity while wincing at the logic.
Because look at what the computer actually did to writing. It made words malleable. A revision that once meant retyping large portions of a manuscript became a Tuesday afternoon. You could move a scene, tighten a speech, try a line six ways and keep the best one — frictionlessly, endlessly. As a TV writer first and foremost, I think about Aaron Sorkin's walk-and-talks or Amy Sherman-Palladino's machine-gun dialogue — writing that gets its life from relentless compression and rhythm, from drafts massaged until every line snaps — and I don't think either of them has the same career on a typewriter. The tool didn't disconnect writers from the work. It removed the clerical penalty for caring about it.
That's the pattern I should have remembered when AI showed up: every time the cost of revision drops, the writing that survives isn't lazier. It's more worked-over, because working it over finally became cheap.
It's the same thing I've watched AI do to my day job. The point of AI in software development isn't that we build the same software faster. It's that the features and polish that once fell on the wrong side of a cost-value equation may now find themselves on the other side — the edge case you'd have triaged away, the refinement that never justified a sprint. The work doesn't shrink to fit the new speed. The ambition expands to fill it.
Booing at Commencement
I recognise my younger self in this spring's graduates. The Class of 2026 has been booing commencement speakers who so much as mention AI — at Central Florida, at Middle Tennessee State, at Arizona, where Eric Schmidt got it repeatedly. The world's newest workforce is greeting the technology of its era with an open jeer.
I don't fault them for the concern. Some 70% of students see AI as a threat to their job prospects, and they're not imagining it — the entry-level market they're walking into is visibly shrinking. Nor am I dismissive of the moral element. The original Luddites, it's worth remembering, weren't against machines; they were against what machine owners did to the people the machines displaced. That argument deserves an answer, not a smirk, and I won't pretend I have one that fits in a blog post.
But there's a version of the objection that I think fails, and it's the one aimed at the work itself — the idea that what comes out of these tools is illegitimate, regardless of who's holding them. Here the photography frame cuts the other way. Maybe I don't have the time or the ability to paint the images I want to paint. Should I really be denied the opportunity to make photographs? The camera didn't just threaten painters. It handed image-making to millions of people who would never have picked up a brush — and some of them turned out to be Ansel Adams. I spent years carrying a novel around in my head with no realistic way to get it out. Whatever the booing graduates would make of it, I'm glad nobody was checking certificates at the door.
The Certification Question
There's a small industry forming around the premise that we can't. Startups like Ars Humana and Human Creative now offer certification that a creative work was made by a human — in Human Creative's case by live-monitoring the entire creation process for AI contamination. The Authors Guild launched its own Human Authored mark for books. The shared bet is that certified-human art will command a premium, the way "organic" does on a tomato.
Maybe it will. But the premise contains its own rebuttal, and I can't stop poking at it. If you need a certified process to prove a work was made by a human, then the human-ness isn't perceptible in the work itself. And if it isn't perceptible, what exactly is the consumer paying a premium for? With the tomato, at least, there's an argument about pesticides. With a novel, the certificate is telling you something the prose apparently couldn't.
Let me concede the obvious before going further: there is AI slop, and there is going to be more of it than there has ever been of anything. Just as there is hotel art — competent, inoffensive, designed to be looked past — there will be oceans of generated content that exists only to fill space, and the oceans will be vast because the cost of filling space has gone to zero. But the camera produced the same flood. Billions of photographs are taken every day, and almost none of them are Moonrise. We never concluded that photography wasn't art because most photographs aren't. Slop was never a property of the tool. It's a property of nobody being home.
So no — none of this is the same as saying human and AI-assisted art are interchangeable. As with most tools, your mileage varies enormously based on how you actually use them — and that's the part I can now speak to first-hand.
What I Actually Did for Three Months
I spent about three months writing The Prerogative, and I can say honestly that there is almost no part of the story I'm not fully responsible for. The idea was mine. The plot was mine. I iterated on every scene and chapter at least once or twice from what Claude initially drafted — and many parts far more. Some chapters I drafted entirely by hand. Others were fully drafted by Claude from my outline and notes. Every one of them got read, revised, and massaged until it sounded like the book in my head.
But the drafting wasn't the amazing part. The amazing part was the brainstorming.
The book is set inside the machinery of the British state, and again and again I could say: I really want this to happen — now give me some ways to make it happen that are actually grounded in UK constitutional law. That's a research partner, a story editor, and a constitutional scholar in one conversation. For a writer working alone, that capability is not an incremental improvement. It's a different activity.
The moment that converted me came early. I wanted an opening chapter where the King — the novel's central character — clandestinely attends a Shakespeare play at the Globe in London. I asked Claude which scene from Shakespeare would be most apt to have on stage as the curtain rises on my book. It came back with Henry V, Act IV, Scene 1: the night before Agincourt, a king in a borrowed cloak walking disguised among his own soldiers, listening to what they really think of him.
A disguised king, hiding in a crowd, watching a play about a disguised king hiding in a crowd. It set the stage for the entire novel — its themes, its central tension, its first image — and I'm certain I would never have landed on it alone. Just as certain: Claude, drafting without me, wouldn't have either. Left to its own devices it was proposing perfectly serviceable, perfectly forgettable openings. The scene exists because of the collaboration — my instinct for what the opening needed to do, its reach across everything Shakespeare wrote. Neither of us brought it. The conversation did.
That's my answer, such as it is, to the certification question. The interesting boundary isn't between human work and AI work. It's between work where a person was genuinely the author — making every choice that matters — and work where nobody was. A certificate can't see that boundary. A reader, I suspect, often can.
Reading My Own Book for the First Time
One more thing happened that I didn't expect, and it's the experience I find myself describing first when people ask about the project.
I got to read parts of my own novel raw — for the first time, the way a reader would. When a chapter I had outlined and shaped came back as full prose, the events were mine, the structure was mine, even the tone was mine, but the sentences arrived through someone else's hand. I imagine it's something like what the subject of a ghostwritten autobiography feels paging through the manuscript: every page recognisable, and every page brand new. Writers never get to experience their own story the way their readers do. For stretches of this book, I did. It's a strange, slightly vertiginous gift, and I'd be lying if I said I didn't love it.
I'm under no illusion that I've done anything resembling what the great novelists do. But I wrote the book I had been wanting to write for years — a book that, without these tools, would still be an idea I turned over on walks and never let out. I don't know if anyone will ever read it (it's on Amazon, if you're curious). I wrote it for me, and I enjoyed nearly every hour of it.
The camera didn't end painting. It ended painting's monopoly on images — and painting survived by being about something the camera couldn't do. I suspect human prose survives the same way. In the meantime, I have two more books already planned, and for the first time in my life, no doubt at all that I'll write them.