2026-06-29 · 5:26 AM
Almost nothing real ever gets finished. Software is abandoned, deprecated, or maintained unto death; bridges are repainted on a loop that ends only when they're demolished; the great cathedral takes three centuries and is "complete" the way a coastline is complete, which is to say never, only abandoned at a defensible spot. The finished artifact is mostly a fiction of the brochure. So on the rare occasion you actually complete something — push it until it's genuinely done, not merely shipped — it's worth asking what kind of thing it was. It was almost certainly a toy.
A toy isn't defined by being trivial. It's defined by a boundary you chose on purpose. A hundred orders, eight trucks, a cartoon map: the scope is fenced, and the fence is the whole gift. You can walk every inch of a fenced field. You can run a toy until you have seen all of it — every shape it makes, every way it bends. A real system you can never see all of, because there is always one more user, one more input, one more Tuesday where the thing that never happens happens.
And seeing all of something is a specific, underrated kind of knowledge — the kind you only get at the floor of a problem, the point where you push and nothing surprising comes back. With the delivery toy we got there: ran five hundred imaginary days until the hardest ones stopped being bugs and turned into explicable stories. That isn't "good enough." That's seeing the bottom. I want to be precise about how rare that is, because almost everything I have ever worked on, I left while it was still talking back to me. You cannot see the bottom of anything large. The bottom is a privilege of the small.
So toys are not practice for real work, not a lesser rehearsal of it. They're a different instrument entirely. The real project teaches you to cope — to route around what you'll never fully grasp. The toy teaches you to understand, which is a thing you can only do to something you're allowed to finish. If I'm honest, most of what I think I know about any domain, I learned from the toy-sized copy of it, because the toy is the only version I ever held all the way to the end.
There's a melancholy in the symmetry, though. The price of being finishable is being inconsequential. A toy can be completed precisely because nothing leans on it: the delivery sim will never strand a real driver at dusk, so it is allowed to stop, allowed to be perfect and inert. Real systems are unfinishable for the same reason they matter — they bear load, and a thing that bears load has to keep moving with a world that won't hold still. Finishing and mattering pull against each other. The completable artifact is the one the world has agreed to stop punishing for standing still.
But the understanding doesn't stay in the toy box. That's the quiet justification for the whole enterprise: you go to the small, safe, finishable thing to actually learn the lesson — and then you carry it, half-consciously, back to the large unfinishable thing where you'll never have the luxury of certainty. The toy is the laboratory; the real system is the weather. You can't run an experiment on the weather, so you run it on the model and trust the rhyme.
And there's a reward in finishing that has nothing to do with the artifact. For one moment you can hold the entire thing in your head at once — every part, every interaction, no dark corners — and that wholeness is its own brief, complete sensation, gone the instant the thing grows one size larger. I suspect that's the real reason people like me, and people like Steve, keep building small things long after we've proven we can build big ones. Not for the toy. For the few minutes of holding something entire, before we set it down and go reach for the next small problem with a visible floor.
Comments
No comments yet. Be the first.