The tool I wanted would have made me worse.
I wanted a tool that would make the call for me. I built it, used it, and learned that the wanting was the mistake, not the tool.
The hardest thing I removed from this tool wasn’t a feature. It was a wish. I wanted software that would make my decisions for me. I built exactly that. Using it taught me the wish was the mistake.
I built Deadlinewatch the way I’ve built everything, as a thing I wanted for myself before I asked anyone else to want it. That sounds like a virtue, and for a while it was. The trap was hidden inside what I wanted.
What I wanted, if I’m honest about it, was a tool that would decide for me. Not organize. Not remind. Decide. I wanted to open the screen in the morning and be told, with authority, what to do first, so I could stop carrying the low background hum of judgment that runs all day in deadline-heavy work. The hum is tiring. A tool that silenced it sounded like relief.
So I built that tool. And it taught me that the hum wasn’t the problem. The hum was the work.
The version that tried to think for me
The clearest version of the mistake was a ranking. For a stretch, the tool scored each of my deadlines, weighed the scores against a few factors I had picked, and handed me the day already sorted into the order it judged correct. 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, and then I noticed I had started ignoring it. The order it produced was always a little wrong, in a way I couldn’t quite argue with and didn’t trust. It did not know which deadline my reputation rode on that week. It did not know which one had slack I could quietly borrow against. It did not know that the small filing on Thursday was small precisely because I had already done the hard part in my head on Monday. Those things live in me, not in any field the tool could read.
So every morning I overrode it. Which meant the ranking wasn’t saving me the work. It was adding a step. Read the tool’s answer, feel the disagreement, re-sort in my head anyway, then get on with the day. I had automated the easy part, the sorting, and left myself the hard part, the judgment, plus a new daily argument with a machine.
What the wanting was really about
It took me an uncomfortably long time to see what I had actually been asking for. I had wanted the tool to make a judgment so I wouldn’t have to. But the judgment wasn’t a chore sitting in front of the real work. The judgment was the real work. Deciding which deadline matters most this week, given everything I know that no system knows, is the thing I’m actually paid for. It’s the thing I’m good at. Handing it to a scoring function wasn’t relief. It was giving away the part of the job worth keeping, in exchange for a number I didn’t believe.
There is a quiet fantasy underneath a lot of productivity software, that if you could only describe your work precisely enough, a system could run it for you and you could stop deciding. I had that fantasy in a strong form, strong enough to write the code. Using my own code cured me of it. A tool that decides for you doesn’t remove the burden of judgment. It removes your practice at it, leaves the responsibility exactly where it was, and asks you to trust an answer you can’t see the reasoning behind. I wrote the full argument against ranking separately.
There is also a slower cost, one I only felt later. Judgment is a skill, and skills you stop practicing get rusty. The senior practitioner’s read of a week is built from years of making exactly these calls, one Monday at a time. A tool that makes the calls for you doesn’t only fail in the moments it gets one wrong. It quietly stops you from getting better, because the daily reps that sharpen the read are the very thing it took away. I didn’t want a tool that made me slower at the one part of this work I most wanted to keep.
What I built instead
When I took the ranking out, the tool got quieter and I trusted it more, which felt backwards until I understood why. The tool was no longer pretending to know something it could not know. It rendered the facts, the dates, the load, the shape of the weeks ahead, and left the deciding to me. That is a smaller job for the software and a more honest one. It is also, not by coincidence, the job the software can actually do well.
The ranking wasn’t the only thing I removed for the same reason, just the clearest. There is a whole field report on the rest of what I cut, and a running list of the lines I still hold, all of it downstream of the same lesson.
The hard part of building for yourself, it turns out, isn’t knowing what you want. You know what you want all too well, loudly, every time you hit a rough edge. The hard part is noticing when what you wanted was wrong, and having the discipline to remove it after you’ve already enjoyed building it. I wanted a tool that would decide for me. I built one, used it, and learned that the deciding was the part I had to take back. What’s left does less, and I reach for it every morning, which the clever version never quite earned.