Coding agents are getting incredibly good, and I am having a lot of fun with them. There is something quite addictive about having an idea and being able to follow it much further than you expected. A thing that would have needed a weekend of figuring out unfamiliar tools becomes something you can start now. You make a change, see what happens, make another. I wrote last year about shipping more while becoming worse at programming from scratch, and the discomfort of that trade has stayed interesting to me. But so has the pleasure. I don’t want to forget that part while discussing what we might be losing. Building things had become exciting again, and a lot of that excitement came from being able to get past the places where I would previously have stopped.
Around the same period, I was working on my fitness. Between October 2024 and September 2025, I lost 38 kilograms. I can put that number here, but I don’t really have a neat way to write about the year it took. A before and an after make an experience look much tidier than the time spent inside it, a year folded small enough to fit in a sentence. The result gives you somewhere to look, and everything in between gets filed under whatever explanation you prefer: consistency, discipline, changing your habits. Those words are useful enough, but they can make a year sound like the execution of a decision. There is quite a distance between deciding you would like something to change and continuing to take part in that change. Having a clear idea of the result doesn’t let you stand outside the process and wait for it.
This year I also started learning the drums. I was shit for about four weeks before I started getting the rhythm. Then I plateaued again, broke through again, and now I can play basic grooves. That is the extent of my expertise, so please take the musical part of this essay accordingly. The second plateau is worth mentioning because it spoils the lovely story where you persevere through the beginning and everything becomes easier. I had got somewhere, and then I was stuck again. There was improvement, but it wasn’t a reliable little payment for each bit of effort. I can describe the progression now because I know what came after it. During those first weeks, all I had was what I could manage at the time.
I find it easy to be generous about that sort of beginning when someone else is doing it. Of course you can’t play yet. You’ve only just started. It seems reasonable that learning an instrument should take a while, and there is no shortage of encouragement available to a beginner. But much of that encouragement comes with a promise: keep going and you’ll get good. I understand the appeal. I want to get better too. Still, I’d like there to be a little room for the possibility that I do this for a while and remain fairly ordinary at it. Basic grooves already give me something to play, a small patch of music I can find my way around. There is plenty beyond it, but I don’t have to spend the whole time looking over there. I can enjoy what is available to me now and still want to learn the next thing.
Habit is interesting here because it can make room for an attempt that doesn’t amount to very much. You return to something because it has become a familiar corner of the day, without having to assemble a convincing argument for it each time. I like that more than the heroic version of discipline, where every repetition proves how badly you want the result. Sometimes the result is too far away to be much use. Sometimes you don’t know how much you want it yet. You need a little time with the activity before you can tell, long enough for some of the strangeness to wear off. If every awkward attempt becomes an occasion to reconsider whether you have any aptitude, it seems quite possible to talk yourself out of something you would eventually have enjoyed.
There is a similar awkwardness in writing. A sentence can be wrong in a way that takes some time to understand. You might have used the wrong word, or you might discover that you don’t quite believe the thing you are trying to say. For this piece, I asked an AI to help me think, and then to help me write. The first outline became a defence of being mediocre. I wanted something quieter, with room for habit and flow, so we changed it. Then the draft arrived and each paragraph seemed to know exactly why it was there. Fitness illustrated one point, drums illustrated another, and the whole thing moved efficiently toward an ending. I asked for longer paragraphs and said I was happy to digress. I wanted a little air between the thoughts, somewhere for a sentence to go without already knowing what it would find. I had help producing the words, but I was still finding out what I wanted the piece to feel like by reading what felt wrong.
That is part of what makes the question of effort difficult. I can ask for a finished version while I am still working out what I mean by finished. With coding, the first moment of resistance now offers a convenient place to hand the problem over. Much of the time that is useful. I don’t need a personal relationship with every task involved in building something. But if I hand over anything as soon as it starts asking something of me, I might leave before I have had a chance to get interested. There is often a stretch at the beginning where your attention is still elsewhere, as though you’ve sat down but haven’t quite arrived. The details take a little while to become familiar. I wonder how often I mistake that stretch for a verdict on the activity itself. Being able to bypass it so easily makes it worth noticing, even if I choose to bypass it anyway.
And then there is the other experience, which is just as real to me: help can keep the whole thing moving. That was there in my early excitement about coding agents. I could spend longer developing an idea because a task within it no longer took all the time I had. There is a kind of symphony possible in that, with individual parts being worked on while you stay with the shape of the whole. You can notice that something doesn’t fit, change your mind about it, or follow an unexpected possibility without losing the rest of the work. There can be a quiet gathering of attention in that, until the separate tasks begin to feel like one thing you are making. It would be strange for me to celebrate the pleasure of involvement and then insist that it only counts when I do every part myself.
I don’t think these experiences need to be reconciled into a sensible policy about when to use AI. They can sit next to each other for a while. I like what automation makes possible, including the chance to give some things more time than they would otherwise get. I would also like a little of the day to remain unclaimed. There is a temptation to treat every task taken off your hands as a vacancy to fill with another task, until the day is full again and the only visible difference is how much passed through it. I’d be quite happy for some of the difference to be an activity I am still learning, or a piece of writing that wanders a little before it finds somewhere to stop.
The drums are still at basic grooves. There is plenty left to learn, and another plateau would hardly be a surprise. For now, though, I can play a little. I was bad at it for a while, and I’m still new enough to have a lot of that ahead of me. There is room for it. I don’t need to hurry this part along.
The pleasure of being bad.
The pleasure of being human.