Starting small is enough

A first version of almost anything can now be built in an evening. Describe it and a model writes it, or type it yourself and finish before midnight. By bedtime, it works. This is a good thing, and worth doing — a small thing that actually runs beats a month of it living in your head.

Everyone wants it fast and cheap. First, one line is worth keeping straight: is the thing small, or thin? Small means fewer parts — one screen first, the other nine for later. Thin means the same parts, each with work skipped, skipped where it doesn’t show. On the evening it works, small and thin can look identical. The difference surfaces much later.

“Small” sounds like a compromise. It isn’t. A first version has one job: to carry the single most important path from end to end. The other nine screens, the features you’ll only want later — them being absent is fine. They simply haven’t come up yet.

Thin is different. What it skips is the kind of work you only draw on when you change things: the shape of the data, what happens when something breaks. On day one it does nothing, so you can’t see it missing. You only find it gone the day you go to change something.

Concretely. Suppose the skipped corner is a small one: a tax rate, typed by hand into four different files. On day one, nothing looks wrong. The day the rate has to change — it moves, or it was wrong from the start — you find that changing it means knowing there are four, and finding all four. A small first version doesn’t carry that: it simply hasn’t built that part yet, and when it does, it builds it once, properly.

Cheap is a size, not a method.

“Ship it thin, fill it in later” often doesn’t pay, for the same reason. Filling in means making the same thing a second time, and doing it while people are already using the first. Small sidesteps that: there are no holes waiting to be patched, the rest just hasn’t been built yet.

So the instinct to start small is right the whole way through. What you’re protecting is the biggest unknown: whether anyone wants the thing at all. A small first version that works is just enough to answer that. It doesn’t have to grow for that — it just has to actually finish the small part it takes on.

Back to the evening. It ends with something that works. It’s small — that’s fine, small was the point. As long as it’s the small-and-whole kind, what you’re holding is a real start — small, but the whole of what it set out to do. And that’s enough.