Making got cheap. Checking did not.
You can have a website by dinner, and so can everyone you are pitching against. What actually changed, what it means for the platforms, and why your client can now produce something that only looks finished.
August 5, 2026
You can have a website by dinner. Not a wireframe, not a mockup: pages, styles, working forms, deployed. So can your intern, and so can the person you are pitching against.
I should say up front that this site was built that way. I wrote the arguments and made the decisions, and an assistant wrote most of the markup. If that reads as a conflict, it is at least a disclosed one, and it is why I have opinions about where this actually lands.
The interesting part is not that production got fast. It is which part of the job got fast, and which part did not move at all.
Making got cheap. Checking did not.
Producing a page is now close to free. Deciding what the page should do, whether the thing you got is correct, and whether it will still be correct in six months — none of that got cheaper. If anything it got harder, because there is now much more output to check and it all arrives looking finished.
That is the whole shift in one line, and most of the noise in this industry comes from people accounting for only the first half of it.
The work did not disappear. It moved from typing to judgment, and judgment does not have a progress bar.
Here is what that looks like in practice. Building this site, an assistant and I shipped a stylesheet where four section headings were invisible: correct markup, valid CSS, a class that never landed. It shipped. It sat there for two weeks. The page looked fine, because a missing label looks exactly like a design decision.
Elsewhere the same week: dark text on a dark panel, a navigation that quietly failed contrast requirements, and an animation that faded things in and then left them permanently out of position. Every one of those passed a syntax check. Every one of them needed a person to look at the screen and say that is wrong.
What this means for Squarespace, Webflow and the rest
The obvious take is that site builders are finished, because why pay for a template when a model will write you the code. I do not think that is right, and I say that as someone who has built on Webflow for years and self-hosted this site.
What those platforms actually sell is not the drag-and-drop canvas. It is hosting that does not fall over, a content editor a non-technical person can be handed without fear, forms and commerce that already work, and someone to call. None of that is what a model is good at, and all of it is what an organization actually needs on the day after launch.
What is genuinely exposed is the middle: the template marketplace, the agency whose value was assembling a template faster than you could, and the tier of work priced on production hours. If your pitch was that you can build it and the client cannot, that pitch has a shelf life.
The likely shape is that platforms become the place work lands rather than the place it is made. The canvas gets less important. The runtime gets more important.
Your client can now produce something that looks finished
This is the change that will cause the most trouble, and almost nobody is talking about it.
Before, a client attempting their own site produced something visibly rough, and the gap between that and professional work was obvious to them. Now they produce something that looks credible on the first pass. The gap has not closed. It has become invisible.
So the conversation changes. You are no longer being hired because someone cannot make a website. You are being hired because they cannot tell whether the one they made is any good, and they have worked out that this is a different problem.
Which means the things you now have to be able to say out loud are: what will break, what will not survive contact with their staff turnover, what will fail an accessibility audit, what will read as generated to the people they are trying to impress. That is a harder sale than a portfolio. It is also a much better one, because it is not reproducible in an afternoon.
Three parties, and the middle one has no stake
There are three participants in a build now: the person who wants the thing, the person accountable for it, and a model that is extremely capable and completely indifferent to the outcome.
That last part matters more than the capability. A model will produce a confident answer at the same speed whether it is right or wrong. It has no sense of what it will cost you to be wrong in front of a board. It does not know which of your claims you can defend. It will not push back when you ask for something that will embarrass you in six months, unless you have set it up to.
So the accountable person is not doing less work. They are doing different work: deciding what should exist, checking what came back, and carrying the consequence. The typing was never the expensive part. It just used to be the visible part, which is how it ended up in the pricing.
If you are running a small studio
Stop pricing production. Price the judgment, the verification, and the fact that you are still there in a year. Those are the things that did not get cheap.
Get faster at the checking, not only the making. Screenshots over assumptions. Real devices over emulators. Read the thing aloud. Most of the failures I have described were invisible to every automated check and obvious to one person looking at a screen for four seconds.
And be honest with clients about what changed. They will find out anyway, and the ones worth keeping will respect a studio that says the building part got cheaper and here is what you are actually paying for, over one that pretends nothing happened and quietly keeps billing for hours that no longer exist.
