The Draft That Waited Four Years
The Draft That Waited Four Years
Published 14 September 2026 There is a file in this folder called `winter-light.md` and until March I had not opened it since 2022. It is 1,900 words about a walk I took in February of that year, and it stops mid-sentence, in the middle of a clause about the way the light came off a wet road. I know exactly why it stops. I could not work out what the walk was about, and rather than admit that I told myself I would come back to it when I did.
What I found when I opened it was not what I remembered writing. I remembered \
a failure. What is actually there is a perfectly good description of a road in February that simply does not need the paragraph of meaning I was straining toward when I stopped. The essay was finished. I had been trying to add something to it that would have made it worse, could not make that addition work, and mistook the resistance for incompleteness.
"Most of my unfinished pieces are not unfinished. They are finished pieces with one more paragraph attached that I could not write because it should not exist.
"
I have been keeping these essays in the same folder since 2019 and the thing \
that has changed most is not the writing. It is what I do with a draft that stalls. For the first two years a stall meant a problem to be solved, and I would push at it, and about half the time I would push the piece into a shape it did not want. Now a stall means a question: is this a piece that is not done, or a piece that is done and does not know it? The test I have landed on is embarrassingly simple. I read the draft aloud and stop at the last sentence that is actually there. If the ending lands, it was never unfinished. If it trails, it is genuinely short of something. Winter light landed. It took four years and ninety seconds to find that out.
The other thing four years does is make you a stranger to your own sentences. \
I could not have edited that piece in 2022. I was still the person who had taken the walk and I could not tell the difference between what was on the page and what was in my head about the walk. In 2026 I have no memory of the road at all beyond what the file tells me, which turns out to be exactly the position a reader is in, and exactly the position an editor needs. So the file sat for four years and the sitting was the work. That is not a method I can recommend, because it is not a method. But it is worth knowing that the folder is doing something while you are not looking at it.
A note on the archive above
Sixteen pieces over seven years is not a lot and I have made peace with that. The gaps are real: nothing at all between February 2022 and May 2022, and only one piece in the whole of 2025. Both gaps are in the list because taking them out would make the list a lie about how the work actually goes.