I thought this was the part you hide

None of it seemed like anything to show anyone — the implementation notes, the dead ends, the calls I got wrong. Show it once it works. Nobody benefits from watching you get stuck halfway. Underneath that, I'd decided failure was embarrassing without ever quite putting it in those words. A record meant the finished parts, laid out neatly.

Office work runs on the same instinct, and there it's fine. A proposal circulates; an ugly draft doesn't. You look over an expense report once before it goes out, instead of filing it in the handwriting you scribbled the numbers in. And the looking-over happens because somebody at the other end is going to read it. That last part is what I missed for a long time. The record of an implementation could have worked the same way, and instead I'd filed it under things you only show finished.

What changed it wasn't much of a decision. I put something out for no particular reason. Nobody suggested it. If there was a motive, it was a weak one: hiding had started to feel like more work than not hiding.

So the notes went out with the failed part still sitting in them, and something I hadn't expected happened. Writing took longer. Deciding got faster.

The reason turns out to be dull. Writing down "why I chose this" forces the choice to be legible, and the parts of it that were vague don't survive the trip. Sometimes what surfaces is a call that had thinner grounds than I'd assumed. Sometimes it's the memory of getting stuck in this exact place before and forgetting about it.

The one I caught most often: I'd pick an approach first and write the reasons afterwards. The honest account fit on one line — I tried it, it worked out, I kept it. But by the time I was writing it up, a respectable-sounding rationale had already moved out in front, and I'd have sworn that was the order things happened in. Putting it in sentences pulls the two orders apart, the one I decided in and the one I explain in. The same thing surfaced when I set out to write up a comparison between approaches. I stopped partway through: I hadn't compared anything. I'd taken the first idea that came to me.

Doing it in public sharpens all of that. Writing for someone puts a pull on the thing to hold together, and that pull does the sorting whether or not a reader ever turns up.

Then there's the second one, which took longer to separate out. Being read isn't what does the work. Being readable is — the possibility, sitting there. This isn't about comments or responses. Keeping the record in a form that could be shown is by itself the quality mechanism.

It shows up in small stalls. I go to write down something I decided casually and my hand hangs for a second: does this hold up to an outside read? Stalling means going back through it. Sometimes the call changes. Sometimes a sentence of explanation is all it needed. Either way the sloppiness comes down without anyone intervening.

One that genuinely stopped me: I'd set the threshold on a check at "that'll do", and started writing the line that justified it. Nothing came. I couldn't say why the line sat where it sat, so I went back into it. Unpublished, that number stays exactly where it is, I think.

That habit has changed how I hand work to the AI agents, too. It used to be that an instruction went out and that was the end of it; what the agent had been working from afterwards was nowhere to be found. Now I write assuming another person, or another agent, reads it back later. The instructions look different — conditions and criteria written out before anything gets handed over, concrete lines where I used to say roughly like this. I think the share of work that comes back off-target has dropped. It feels much the same with people.

So the stuck parts stay in. Two reasons, and both of them are selfish: writing it out sorts out what I decided, and the possibility of being read keeps the carelessness down. There's a third, thrown in — a stuck passage probably helps somebody more than a smooth one does, since a finished result doesn't show you where it caught. Good if that's true. It isn't the part that's actually working on me.

I'll get stuck the same way again. The wrong calls and the do-overs go in as well, written down plainly.