The architecture isn't mine

Status Trace exists because of AI. It's also why I feel less connected to it than to anything else I've built.

I built it fast. A weekend, less than that in real work. It's the kind of thing I'd never have started otherwise. The cost of building it used to be high enough that the idea would have stayed an idea. AI lowered that cost to almost nothing, so the thing got made. I like that it exists. I use it.

But when I look at it, I don't feel what I used to feel.

The old satisfaction was never in having a working product. It was in the structure. Knowing why the pieces sat where they did. Being able to defend every decision because I'd made each one slowly and rejected the alternatives on purpose. That confidence, the sense that the approach was sound and that the soundness was mine, is the part I actually built for. It's not in Status Trace.

I let AI make most of the design and architecture decisions. Not laziness exactly. It was faster, the results were fine, and at every single step it was easier to accept the suggestion than to sit and derive my own. A hundred small deferrals, and at the end I had a system I could operate but hadn't authored.

The obvious reading is that I lost something. Cleaner to say I made a bad trade. But that's too neat.

Because the alternative wasn't me building Status Trace by hand and feeling proud. The alternative was Status Trace not existing. I wouldn't have paid the old price for it. So the choice was never author it myself or let the machine author it. It was let the machine author it, or don't have it at all.

What I miss isn't in the code. It's the doing. The pride I keep reaching for was pride in the process, not in the artifact. AI kept the artifact and removed the process. And I keep looking at the artifact expecting to find the process there, and it isn't, because it never lived there in the first place.

What I've become on these projects is a product owner. I know how to add features. I can get it running, debug it, extend it, ship it. I understand what it does at the level of behaviour. I don't understand it at the level of structure, and more to the point, the structure doesn't sound like me. Read code I wrote by hand and you'd find my habits in it: the way I name things, the shapes I reach for, the things I over-engineer and the things I skip. Status Trace has none of that. It's competent and anonymous.

That's the part that's hard to sit with. Not that it's worse. It's probably better than what I'd have made alone. It's that it isn't mine in the way that matters to me.

I haven't stopped, and I won't. Hand-building architecture out of nostalgia when the machine does it faster, and often better, isn't a principle. It's a sulk. rcordr is the counterexample. I built that before AI, I can feel the structure in it, and I can't manufacture that feeling now by refusing the tools.

What I think is actually true is smaller than a rule. There are two things I've been calling building, and they've come apart. One is making a thing exist. The other is authoring its structure well enough that it reads as an extension of how I think. AI is very good at the first and takes the second away as a side effect, quietly, one accepted suggestion at a time.

I don't yet know which projects deserve the second kind. Status Trace didn't. It needed to exist, and now it does. But I notice I'm keeping the ones I built by hand a little closer, and I don't fully trust the reason. Maybe it's just that they feel like me. Maybe that's reason enough to keep a few of them out of the tools' reach, even when the tools would win.