Why good products still take time

AI did something real to the cost of building. What took a sprint now takes a weekend. What took a team now takes one person and a model. The scaffolding, the first draft, the whole working prototype — faster, cheaper, or nearly free.

What didn't change is the work that matters.

Typing was never the slow part

Watch where the hours actually go in a product people keep using, and almost none of them are in the typing. They're in the deciding.

Should this feature exist at all? What happens the second time someone opens it — when the novelty is gone and they're just trying to get something done? Which of these five reasonable options is the one we'll still respect in a year? What do we cut so the thing that actually matters lands, without everything else diluting it?

That work doesn't produce a diff. An afternoon spent realizing a feature shouldn't exist leaves nothing to show. But it is the work — the part that separates a product people return to from one they abandon after the first frustration.

A model will generate all five versions of the screen, fluently, in the time it once took to build one. It will not tell you which is worth showing.

AI made the building cheap. It didn't make the deciding any easier. Deciding was always the job.

What running doesn't mean

It is now cheap to ship a product that works. Landing page, dashboard, storefront, booking flow, workflow tool — assembled in a day. The output is real. It runs.

What AI cannot do is make that product ready for someone else to own.

Ready to own is a different standard. The empty state tells you something useful, not just nothing. The error message names what went wrong and what to do. The second screen is as considered as the first. The thing works reliably on the fifth use, not just the demo. The edge case got handled — not because it was in the spec, but because someone cared enough to picture the person who'd hit it.

None of that is hard to generate. All of it is hard to think of. The gap between something that demos well and something worth owning lives entirely in this invisible work. You only notice it when it's missing. Whoever you hand the product to will notice immediately.

Restraint is the expensive part

When building gets cheap, the pull is to build more — more features, more surface, more options. Every addition costs almost nothing to make, and something real for the next person to navigate. Abundance makes restraint harder and more valuable at the same time.

Fast to make and good to use were never the same thing. For a while, careful and slow correlated enough to be confused. AI broke the correlation. Now anyone can ship something that looks finished by Friday.

Whether it's actually ready to hand to someone — a customer, an operator, a founder who wants to put their name on it and launch — that judgment is still entirely human.

The work moved

This is what changes when building gets cheap. Not the amount of work. The location of it.

From typing to choosing. From building to finishing. From assembling something that runs to making something worth handing over.

The boilerplate writes itself now. The work that was always the point still doesn't.