The tool I wanted to use

I built Deadlinewatch the way I have built everything: as a thing I wanted for myself before I asked anyone else to want it. That sounds like a virtue. For a long time it was a trap.

When you build for yourself, you add. Every time you hit a rough edge in your own week, you reach for the code and smooth it. A field here, a setting there, a clever calculation that handles the case you ran into last Tuesday. The early versions of this tool were thick with that kind of cleverness. They could do a great deal, and I was proud of how much they could do.

They were also worse than what I have now.

The version that tried to think for me

The clearest example. For a while, the tool ranked your deadlines for you. It scored each one, weighed the score against a few factors, and handed you the day already sorted into the order it thought you should work. I built it because it was the obvious thing to build. A deadline tool should tell you what’s most urgent. Of course it should.

I used it for a few weeks, then noticed I had started ignoring it. The order it produced was always a little wrong, in a way I couldn’t argue with and didn’t trust. It didn’t know which deadline my reputation rode on this week, or which one had slack I could borrow against, because those things live in my head and not in any field. So I overrode it every morning, which meant the ranking wasn’t saving me work. It was adding a step. Read the tool’s answer, disagree, re-sort in my head anyway.

I wrote the full argument against ranking separately. The short version is that a tool which ranks your work claims an expertise it does not have, and the first time it’s wrong, you stop trusting it. I took the ranking out. The tool got better.

Subtraction was the work

Once I saw it, I saw it everywhere. The graded importance field with its three levels became a single Pin. The hidden Snooze that lost deadlines became a ghost row you can still see. The clever calculations became one honest number: days until due. Every removal made the tool quieter and more trustworthy. The list of what it deliberately doesn’t do is on the methodology page now, and it’s longer than the list of what it does.

The hard part of building for yourself, it turns out, is not knowing what you want. It’s noticing when what you wanted was wrong. I wanted a tool that would tell me what to do. I built one, and it taught me that the telling was the part I had to take back out. What’s left does less, and I reach for it every morning, which the clever version never quite earned.