Photo by Tomas Heligr-Pyke · Bokeh
I am in a weird situation, and I’ve had versions of this conversation with a couple of people now, so I might as well write about it.
AI is a great tool. I use it constantly. It has helped me build things I would struggle to find the time to build otherwise, including the increasingly ridiculous game that grew around Juniper. I can look at what I’ve made and see the progress.
Then I look at myself and don’t really feel like I’ve achieved anything. Some sort of weird parasite, sitting there getting the machine to do things while I claim I’ve been busy.
I don’t think that is a fair description of what I do. It is, however, the feeling I’ve been having, and knowing a feeling is a bit stupid doesn’t seem to make it go away.
I know how it works
This would be a simpler conversation if I had no idea what I was building. There would be an obvious problem to deal with: learn enough to understand the thing before relying on it.
But I do understand the work. I know how the things I build are put together, and I keep checking that I can still do the underlying jobs by hand. Time and time again I prove to myself that the capability is there. Somehow that doesn’t produce the satisfaction I expect it to.
Building something manually has a feeling attached to it that I don’t get to the same degree from using AI. I can be pleased that a feature works and still feel oddly removed from having made it. The result is right there, so it isn’t as though I’m wondering whether anything happened. It just doesn’t land in the same way.
Juniper is a fairly obvious example. The project has grown from an AI character in a small world into this much larger settlement game. I have opinions about how it should work. I make the decisions. I know when something doesn’t fit what I wanted. I am enjoying making it, and I can still finish a stretch of work thinking I haven’t really done anything.
Those two feelings apparently get to exist at the same time. Very helpful.
Doing it all by hand doesn’t fit
The annoying answer would be to stop using AI and write everything myself. I could do more of that. With the amount of work I have, though, the extra time has to come from somewhere, and I don’t have an unlimited supply of it.
For me, doing all of this manually would be too expensive in time. Things would take longer, other work would wait, and some ideas probably wouldn’t get built. That isn’t a statement that nobody should code by hand any more. It is the practical situation I’ve put myself in with the number of things I’m trying to do.
And yes, I can hear how much of this is my own fault. I wanted a world for Juniper and ended up thinking about a multiplayer frontier with railways. I am not exactly helping the time problem.
It feels a bit like a snake eating itself. AI gives me the ability to make more, I find more things I want to make, and then there is even less chance of doing all of them by hand. The tool saves time and I promptly find a use for every bit of it.
Meanwhile the part I used to get satisfaction from occupies less of the day. I don’t know that the answer is to deliberately make every job slower, but I can see why simply finishing more jobs hasn’t fixed the feeling either.
Would I feel this way if I hired someone?
That is the question I keep coming back to. If I employed somebody to help build these things, would I sit there feeling like I hadn’t done anything because their hands were on the keyboard?
Maybe I would. I haven’t worked that one out. But the comparison does make me question whether I am being reasonable about my own contribution. I wouldn’t normally say that somebody running a project has no involvement in it because other people helped carry out the work.
There are still decisions to make, things to understand and results to take responsibility for. I don’t suddenly get to stop caring about whether the thing works because an AI wrote part of it. The work is still mine to deal with afterwards.
Intellectually, that makes sense to me. Emotionally, I seem to want the proof that comes from having personally worked through more of it. I can explain why using help is sensible and then immediately judge myself for using it. That is a fairly exhausting standard to keep applying.
I also don’t want to solve this by pretending every prompt is a monumental creative achievement. Some of what I ask AI to do is ordinary work I would rather have done quickly. It is fine for that to be ordinary. I just don’t want the whole day to start feeling as though none of my own effort mattered.
I haven’t got a neat answer
I probably am not doing anything terribly wrong. I am using a tool, building things I care about and trying to make the available time stretch. There is value in those things even when I don’t happen to feel it.
Still, I miss more of the satisfaction of doing the work myself. Admitting that doesn’t mean I want to throw away the tools or go back and manually rewrite a perfectly functioning project to prove a point.
What I need to work out is which parts I actually want to keep doing with my own hands. There is only so much time in the day, but I would like some of the time I spend making things to feel as good as the finished thing looks. If another week of building Juniper leaves me convinced I’ve done nothing, adding another ten features probably isn’t going to answer that.