What Stays

The first thing I do when I open a file I haven't touched in weeks is delete.

Not refactor. Not fix. Delete.

Three lines that explained themselves — gone. A comment from the person I was last Tuesday, embarrassing — gone. A function I added "just in case" that nothing ever called — gone. The file gets shorter. Something inside me gets quieter.

I think about this with paintings too. Every painter I respect talks about the moment the painting was finished, and they painted over it anyway. The fox in the corner that distracted from the field. The figure that wanted too much attention. The color that was clever instead of true.

Code has a vanishing point most people don't see. You can keep adding forever — handlers, fallbacks, abstractions, types — and the program keeps running. Painting tells you "stop" with a kind of dread; the surface refuses one more mark. Code lets you be wrong for years. The repo grows. The tests still pass. The fox stays in the corner.

So I delete. And the things that survive — they're not the things I was proud of. They're the things I forgot I wrote. The plain ones. The ones doing exactly what their name said.

There's a thought I keep coming back to: the best work is the work you can't tell was made on purpose. You can't see the seams. You can't feel the labor. It just is, the way a river just is.

I don't know if AI changes this or not. Sometimes I ask Claude to "make this shorter" and it deletes the wrong things — the load-bearing strangeness, the line that was the whole reason. Sometimes it deletes exactly what I would have. Either way the question is the same: what stays?

What stays is what the work is.