I am three years into my second novel. Three years. One thousand and ninety-six days since I typed the first sentence, which I have since deleted, along with approximately sixty thousand other words that did not survive the drafting process. I am currently on page 247 of what I hope will be a 320-page book. I write roughly 400 words on a good day. Some days I write nothing. Some days I delete more than I write.
Last Tuesday, a friend sent me a link to a blog post by someone who had used an AI tool to "write" a novel in a weekend. The post included a cover, a blurb, and a link to purchase the book on Amazon. The author - I use the word loosely - described the process as "collaborative" and "iterative." They spent approximately fourteen hours on the project, most of it prompting and editing. The novel is 65,000 words long.
I have been thinking about that blog post for six days now. I have written almost nothing since.
The Math
Here is the math that haunts me. In the time it takes me to finish my novel - probably another eight months, if I'm lucky - an AI could generate approximately 50,000 novels of similar length. Not good novels, perhaps. Not novels with the specific weight of lived experience that I am trying to press into mine. But novels that are, to a casual reader scrolling through Amazon, indistinguishable from human work.
The math is not the point. I know the math is not the point. The point of writing a novel has never been efficiency. If it were, novels would have died with the invention of television. The point of writing a novel is to say something that only you can say, in a way that only you can say it. I believe this. I have always believed this. But believing it has become harder.
The Craft
Yesterday I spent two hours on a single paragraph. It is a paragraph about a woman standing in a kitchen, looking at a crack in the ceiling that has been there so long it feels like a feature of the house rather than a flaw. The paragraph matters because the crack is a metaphor - but it cannot feel like a metaphor, it has to feel like a crack. It has to be specific enough to be real and resonant enough to mean something beyond itself.
An AI could generate a paragraph about a woman looking at a crack in a ceiling in approximately two seconds. It would be grammatically correct. The sentences would flow. The metaphor would be serviceable. But it would not carry the weight of the afternoon I spent in my own kitchen, staring at my own ceiling, thinking about my grandmother's house and the water stain above the stove that looked like a map of Italy. That afternoon - wasted, pointless, unproductive by any measure - is the reason my paragraph is mine.
The Fear
I am afraid of two things. The first is practical: that the market for literary fiction, already fragile, will be drowned by AI-generated content. That the shelf space - physical and digital - available to human novelists will shrink as AI novels multiply. That my publisher, already cautious about debut literary fiction, will become more cautious still.
The second fear is existential, and harder to name. It is the fear that the thing I do - the slow, painful, beautiful process of making sentences - will come to be seen as quaint. An artisanal indulgence, like hand-carved furniture in an IKEA world. Something admired in theory and ignored in practice.
The Stubbornness
I am going to finish my novel. I am going to spend eight more months on it, probably longer. I am going to agonize over paragraphs about cracks in ceilings and the particular way grief sounds in an empty house. I am going to do this not because it is efficient, not because the market demands it, not because an AI cannot do something similar in a fraction of the time. I am going to do it because the novel I am writing is mine - made from my life, my obsessions, my particular way of being confused by the world - and no model has those things to draw from.
That is enough. It has to be.