There is a page in my notebook, dated last April, that lists three unfinished posts. I know the date because I wrote it in the corner the way you write a date when you are trying to convince yourself the list is temporary. One post was about why I stopped checking analytics. One was about a comment I never answered. One was a long, tangled piece about what fifteen years of the same blog has actually taught me about attention. Rereading that page recently, I noticed something I hadn’t expected: the discomfort wasn’t about the writing. It was about the rereading. Each draft still had a pulse. Each one still sounded like something I wanted to say. They had simply stopped moving.

That distinction — stopped, not dead — is where this essay lives. Because when I finally sat with why those three pieces stalled, and what role a drafting tool might honestly play in unsticking them, I found that most of the conventional advice about both problems is wrong in the same direction. It assumes the problem is production. It is almost never production.

Why drafts actually stall

The standard explanation for a stalled draft is mechanical: you ran out of time, or ideas, or discipline. The hustle-culture version adds a solution that is somehow both vague and demanding — build a content calendar, batch your writing, ship daily. None of it addresses what I think is the real engine of a stalled draft, which is closer to what Seth Godin describes when he writes about modern vanity: the shame attached to being “not enough.” As he puts it, “avoidance has no floor. The threat of not-pretty (or not-competent, not-respected or any other imagined failure) is always right behind you. You can never outrun it permanently.” A draft sitting in a folder is a draft nobody can judge. Publishing it is the moment the judgment becomes possible. So the draft waits, and the waiting feels like prudence when it is mostly fear wearing a cardigan.

Godin’s question is the useful one: “If you didn’t care about blame or credit, what would you ship?” The answer tells you what you’re withholding. When I asked it honestly about my three stalled posts, the answer was all three. They stalled not because I lacked the skill to finish them but because finishing them meant being seen holding a position. That is a different problem, and it needs a different response than a content calendar.

But there is a second reason drafts stall that has nothing to do with shame, and it matters for what follows. Sometimes a draft stalls because it has no spine. The idea is real, the material is real, but the structure is missing — the piece doesn’t know yet whether it’s an argument, a story, or a letter. Blogging, as I’ve written before on this site, is a form of thinking: “When I write a blog post, I’m not just telling you what I know. I’m discovering what I know.” A draft with no spine is a draft where the thinking hasn’t found its sequence yet. That is a structural problem. It is genuinely hard, and it is also — and this is the part that surprised me — the one place where a machine can help without doing anything that feels like cheating.

Two problems, only one of which a tool can touch

Before opening any drafting tool, it’s worth triaging the stalled draft by cause. A draft with no spine is a structure problem. A draft with no you is a voice problem. They can look identical from the outside — both are unfinished — but they need opposite treatments.

A voice problem means the sentences are competent but anonymous. You reread a paragraph and cannot find yourself in it. No tool fixes this, and any tool that promises to is promising to replace you, which is the one thing a personal blog cannot survive. The whole premise of keeping one blog for years is that the voice is the moat. A blog is “your home base — the one piece of internet real estate you actually own,” as this site has argued before, and what makes that real estate worth owning is that it sounds like a specific person. A tool that smooths your sentences into a competent average is not scaffolding. It is demolition with good manners.

A structure problem is different. Here the voice is intact — you could read any paragraph aloud and recognize yourself — but the paragraphs are in the wrong order, or the piece has three possible shapes and hasn’t chosen one. This is where I let a machine touch one of my drafts, and I want to be precise about what that looked like, because the word “used” covers too much ground.

What scaffolding actually means

For the long stalled piece — the one about fifteen years of attention — I tried a drafting tool for structural help only. Specifically, I ran the idea through the Unsloppy AI Book Writer, an AI book writing tool, to see what kind of outline it would propose for the material. I want to be careful here: this is one instance, not a verdict on the tool, and I’m not claiming it improved my writing. What it did was hand me a proposed sequence of sections — a shape — that I then argued with for an hour. I kept two of its headings as questions, deleted the rest, and rewrote everything in my own words. The outline functioned as a set of prompts for my own thinking, not a substitute for it. That, I’ve decided, is the honest definition of scaffolding: the tool proposes, you dispose, and every sentence that reaches the reader is one you wrote and would defend.

Ghostwriting is the mirror image. Ghostwriting is publishing prose you did not write, would not have written, and cannot defend if a reader asks you about it. The line between the two is not fuzzy. It is a single practical test: could you explain, in your own words, why every sentence in the published piece says what it says? If a reader emailed you about a paragraph and you’d have to check what the tool meant, the tool crossed the line.

There’s also an administrative edge to this that’s worth knowing about, even for a small blog. The distinction between AI-generated material and an author’s own contribution is becoming a formal one, not just an ethical intuition. As The Passive Voice has reported, copyright registration now asks writers to disclaim AI material and describe their own contribution, and Amazon KDP requires authors to report AI-generated text, images, and translations. None of that necessarily applies to a personal blog. But the direction is clear: the world increasingly expects writers to know exactly where the machine’s work ends and theirs begins. Practicing that knowledge on a stalled draft is cheaper than learning it under pressure later.

The experiment, if you want it

Here is a bounded version of what I did, small enough to try this week and reversible enough to abandon without ceremony. Pick one stalled draft — just one. Before opening anything, decide which problem it has. Read two paragraphs aloud. If you can’t hear yourself in them, close the draft and be done for the day; that’s a voice problem, and it needs you, not software. If you can hear yourself but the piece has no shape, open a drafting tool and ask it only for structure: an outline, a sequence, a set of section questions. Then close the tool’s prose out of your mind entirely and rewrite every sentence by hand. Publish only what still sounds like you — or shelve it, honestly, if it doesn’t.

To make “still sounds like you” concrete, here is a clearly hypothetical example of the test in action. Suppose the tool proposes a section heading like “Key Benefits of Slow Blogging.” If your natural instinct is to wince — because you would never write “key benefits” about something you consider a practice, not a product — that wince is the test working. You rewrite the heading as a question you actually have (“What does slow cost me that I don’t mind paying?”) and write toward that instead. The tool gave you a slot; you gave it a soul. If, on the other hand, you find yourself keeping the tool’s sentence because it says the thing better than you would — stop. That sentence is not yours, and keeping it is the first small withdrawal from the only account that matters.

The quiet economics of it

Because that is what a small blog actually is: an account of trust, deposited one honest post at a time over years, read by a few hundred or a few thousand people who came for a person. The cost of a ghostwritten post is not a one-time embarrassment. It is a slow erosion of the thing that took a decade to build — the reader’s quiet assumption that the byline means something. No tool is worth that, and no stalled draft is either. The three posts on my notebook page are still unfinished as I write this. One now has a spine, argued over with a machine and rewritten entirely by hand. The other two are waiting, and I’ve stopped pretending the waiting is about time. It’s about nerve. That, at least, is a diagnosis. The machine can’t supply the nerve. Neither, honestly, can the notebook. But between the two of them, one draft moved, and it still sounds like me — which is the only test I’d trust in ten years, and the only one I’d suggest you run this week.