When custom got cheap

For as long as I have been doing this work, the first honest question with a new client has never been "what should the site look like." It has been "what can the budget actually buy." And for years, the honest answer for most small businesses was a template.

Not because templates are wonderful. Because custom was expensive in a way that had nothing to do with greed. A fully custom build meant engineer hours, and engineer hours cost what they cost. So instead of pretending the trade away, I built a framework for it: template, wireframe, component library, or scratch, chosen by goals and budget.

The old math

The framework worked because the trade was real. A template got you live fast and cheap, and you paid later in sameness and workarounds. Wireframes and component libraries bought flexibility at a medium price. Scratch bought exactly what the business needed and nothing else, at a cost most small businesses could not justify. Every tier was honest. You got what you paid for, and mostly what you paid for was hours.

I have walked clients through that framework more times than I can count. It was the most useful thing I owned. And lately it is going soft in my hands, because the hours side of the equation is collapsing.

What flipped

AI-assisted engineering is real now. Not a demo, not a party trick: real code, shipped to production, doing real work. And the practical effect, at the level of outcomes, is blunt. The parts of a custom build that used to eat the calendar, the wiring, the boilerplate, the hundredth contact form, the responsive grid I have built a thousand times, now happen at a speed that would have sounded like a lie not long ago.

Which means the price of fully custom is drifting down toward what template work used to cost. Sit with that, because it rearranges the whole decision. Templates existed as a category because custom hours were scarce. When the hours stop being scarce, the honest advice has to change.

I want to be careful about what I am claiming here. The machine does not design. It does not know your customer, and it does not care whether the site tells the truth. What it does is compress the distance between a decision and a working page. The decisions are still the job. There are just fewer excuses standing between them and the result.

• What got cheap: boilerplate, scaffolding, wiring, the mechanical middle of every build.
• What did not: knowing what to build, and what to refuse to build.
• What did not: taste. The tool has no opinion on whether a page reads official or generic.
• What got relatively more expensive: verification. Checking the work is now the long pole in the tent.

The catch nobody says out loud

Here is the part the excited version of this story skips. AI-assisted coding makes good developers faster at being good. It also makes bad developers faster at being bad. Same tool, same speed, opposite outcomes, and the tool cannot tell the difference.

A developer who never checked their work now ships unchecked work at ten times the volume. The plausible-looking bug, the layout that dies at phone width, the confident claim nobody can back up: those got faster too. Speed is an amplifier. It amplifies whatever process it lands in, including no process at all.

So the bar moves. When anyone can produce a great deal of code quickly, the differentiator stops being production and becomes judgment: what gets built, what gets cut, and what gets caught before a visitor ever sees it. The gates around the work, the review, the checking, the refusal to ship something because it merely compiles, matter more now than they ever did when code was slow.

The new advice

What I tell businesses now is a different sentence than the one I used for years. Fully custom is on the table at budgets where it never was. If a shop is still quoting custom like the hours are what they were, ask why. And if a shop is quoting custom suspiciously fast and suspiciously cheap, ask a harder question: who is holding the quality bar, and how would you know.

The framework survives, but the tiers are sliding toward each other, and I think the template tier has the most to lose. Cheap and fast was its entire argument. It just stopped being the only cheap and fast thing in the room.

What does not slide is the rule underneath all of it. The tool was never the product. The judgment was. That was true when the tool was a page builder, and it is true now that the tool can write code. The economics flipped. The bar did not.