The tools that make this blog possible are the same ones I could use to scale it out of recognition. One prompt, ten posts, a publishing schedule that never misses. I have run that experiment. The posts were fine — genuinely fine, technically competent, topically reasonable. No one remembered any of them, including me.
That last part is the tell. If the author can't remember what they wrote, the reader certainly won't.
Slow is the product
What I want from this site is not volume. It's a small, reliable signal — one essay per week, from a real person, thought about for longer than it takes to read. That is the thing I am offering. If I use AI to multiply the output, I change what the thing is. It becomes something else: a content operation, a feed, a publish-or-perish ticker. Those things have their place. This is not one of them.
The choice to publish slowly is a product decision. The product is the essay, and the essay requires that I have actually thought about something, not just that I have arranged words competently around it.
The AI helps me structure an argument I've already had. It helps me find the shorter version of a paragraph I've already written. It helps me catch the places where I've been vague because I haven't finished thinking yet. All of that is valuable and genuinely accelerates the work. None of it replaces the thinking that has to happen before I open the tool.
Writing is thinking. I use the model for everything except the thinking.
The thinking part
There is a more honest answer underneath the editorial one. Writing slowly is how I figure out what I actually believe. The act of trying to articulate something — not describe it, but argue it, commit to a position about it — reveals gaps I didn't know were there.
When I sit down with a topic I think I understand, I often discover that I understand it less than I assumed. The vague places in the draft are maps to the vague places in my thinking. The model can identify them. It cannot fill them. I have to go away and read, or think, or talk to someone, or simply live with the question until the gap closes.
That process takes time. It cannot be compressed past a certain point. Weekly is roughly the pace at which I can do it honestly.
What AI-assisted writing actually means
The phrase "AI-assisted writing" is doing a lot of work in current conversation, and it covers a wide range of very different practices.
At one end: a human who has thought deeply about a subject uses the model to edit, sharpen, and finalize something they wrote. The ideas, the argument, the specific claims — all human. The model touched the prose.
At the other end: the model generates an outline, writes the sections, and the human curates the result. The ideas are the model's. The human made selections.
Both can produce readable output. They are not the same thing. I try to stay closer to the first. Not out of nostalgia or purity — the model's editing is genuinely good and I rely on it — but because the thing I am trying to produce is my thinking, and my thinking cannot be delegated.
The practical upside
There is a practical upside to slow that is often missed. Readers who find a site that publishes once a week and has never missed an issue experience something different from readers who find a site that publishes ten times a week. They trust it more. They read more carefully. They remember more.
Attention is finite. When a publisher signals that they are careful with it — through pacing, through editing, through the sheer fact of scarcity — the reader brings more attention to the exchange. That is the deal I am trying to make: I will be careful with your time if you will give me some of it.
So yes — I use AI every day. And yes — I publish slowly. Both sentences are true, and one makes the other possible.