Last week I did the same setup by hand for what must have been the fifth time, and somewhere in the middle of it I got annoyed enough to stop and ask why I was still doing it at all.

It wasn't hard work. That was the problem. It was a sequence of small, knowable steps — pull the state, sort it, figure out what mattered, lay it out, hand it off — and I knew every step cold. Which is exactly why I kept doing it by hand: it never felt worth the interruption to fix, because each individual time was only a few minutes and I already knew how. So I paid the few minutes, again, and quietly relied on myself to not miss a step while I was tired or rushed.

That reliance is the bug. Not the minutes. The minutes are annoying but survivable. The bug is that the correctness of the thing lived in my head — in my remembering to do all the steps, in the right order, every time, on a bad day as well as a good one. I'd built a process whose only guardrail was my own attention, and attention is the least reliable component I own.

There's a line I keep coming back to now: if the only thing keeping a task correct is that I remember how to do it, the task is already broken. It just hasn't failed yet.

So instead of doing the setup a sixth time, I spent an afternoon turning it into something that does itself — the steps written down as code that runs the same way whether I'm sharp or half-asleep, with a check at the end that refuses to call the thing "done" until the actual conditions are met, not just until I've clicked through the motions. It cost more than the few minutes would have. It cost an afternoon. And it will pay that back within a couple of weeks and then keep paying, but honestly the time isn't even the point.

The point is what moved. Before, the knowledge of how to do it right lived in me, and every run was a fresh chance for me to be the weak link. After, the knowledge lives in the tool, and I get promoted out of the loop. I stop being the person who has to remember and become the person who decides whether the result is good. Those are different jobs, and only one of them scales.

This turns out to be the actual dividing line in how I judge my own work now, and it's not the one I was taught. I was taught to value knowing how — the engineer who can do the thing, fast, correctly, from memory. That skill is real and it's also a trap, because the better you are at doing something by hand, the more invisible its cost becomes, and the longer you'll go on quietly being the single point of failure for it. Being good at the manual version is what hides the fact that it should never have been manual.

The tasks I actually did well last week weren't the ones I concentrated hardest on. Concentration is just attention again, and attention runs out. The ones I did well were the ones I'd made impossible to do badly — where the correct sequence was baked in and the finish line had a real check on it, so there was nothing left for me to forget. When I trust a piece of my work now, it's almost never because I trust myself to be careful with it. It's because I've arranged things so I don't have to be.

And there's a second reason this matters more than the time it saves, which took me longer to see. Every step you keep doing by hand is a step you can't hand to anyone — or anything — else. It's stuck to you, because the instructions for it only exist in your head. The moment you write it down as something that runs on its own, it stops being yours. A teammate can trigger it. A tool can run it while you sleep. You've converted a private skill into shared infrastructure, and the difference between a person who hoards skills and a person who leaves working systems behind them is mostly just this habit, repeated a few hundred times.

None of this is a productivity trick. It's closer to a stance: treat every "I'll just do it by hand again" as a small confession that a system is missing, and treat your own memory as the thing you're trying to design out, not lean on. The goal isn't to be the most reliable person in the room. It's to build things reliable enough that they don't need you to be.

Part of an ongoing set of notes on treating your work — and yourself — as a system worth examining.