A CEO whose company I’ve followed for a while posted a thesis this week, and I’ve been turning it over since. Websites, he argued, are going the way of contact lenses. I’ll paraphrase his point below, but read it for yourself. I left a comment right away, and as I thought about it more, my comment became the rest of what is on this page. The first contact lenses were expensive and durable. You bought a pair, you cleaned them, you kept them in a case, and you expected them to last. The cost obliged a kind of care. Then the materials improved and production got cheap, and the whole relationship inverted. Now you buy a year of them in a box and throw one away at the end of every day. His claim is that AI does this to websites. Spin one up, use it, discard it, make another. Iterate fast enough that lifespan stops being a concern. I think he’s partially right, and, right now, it’s the more interesting part — that’s what makes it feel like it’s the whole. When production costs drop, the obvious effect is that output gets cheaper and faster. The less obvious effect, and the more important one, is that things become possible that were never attempted before. Not because nobody wanted them, but because they could never justify the floor. A landing page for a two-week experiment. A microsite for one event. A version of an idea built just far enough to find out it was wrong. All of that was priced out of existence until very recently. A long tail of short-lived, exploratory things is going to appear where there was nothing before, and some of it will be genuinely good. Disposable describes them to a point, but just because something can be disposed of doesn’t mean it cannot also last. I suspect that the best of the long tail will last, and grow, and make the tail even longer. The Head of the Long Tail But a tail has a head, and I happen to be standing right on it. My firm is redesigning and replatforming an enterprise website right now. It has tens of thousands of unique URLs. It depends upon dozens of data integrations. It is a part of a cluster of subdomains doing different things. It has international properties that live on separate infrastructure in other countries because regulation requires it. It contains thousands of pages of product and industry compliance information that has been through legal review and will have to go through it again. Automation will absolutely reduce the cost of the production involved. It already is. What it will not reduce is everything else, and everything else is most of it. In a project of this complexity, the challenge is decision-making. AI can help with that, too, of course. It can index all the website’s content, identify problems, recommend solutions, and organize objectives. A person has to steer it that way, and people have to react. The decisions still land on people. On this project, there are dozens of stakeholders among several organizations spread across six or seven time zones, which means the work of gathering consensus, sequencing decisions, and unifying expectations costs exactly what it has always cost. That part hasn’t moved at all. Rebuilt by Hand Here is a specific example of where the head — a project like this — eats its own tail. We needed to present the brand direction to stakeholders. I worked with the principal designer responsible for the identity work to extend what had been created into a systematized visual language. We established that system in the canvas, and then automated the assembly of a brand presentation deck using a brand system skill. The deck came together almost instantly and was updated several times up until the last possible moment very easily. The presentation went well. We received feedback both on the deck itself — which needed to be updated and circulated — and on the elements of the brand system it presented. And that’s when the choice to generate a deck rather than build one in a collaborative environment became costly. Generative work issues from a single account. In this case, there are three separate companies collaborating, so even a shared team account wouldn’t have helped. The prompt is just not as collaborative as the canvas. What we ended up doing was rebuilding every slide by hand in Figma, not because the generated deck was worse, but because it could only be touched by one person at a time. The canvas is still the thing: A single surface that several people can stand around and work on at once was an enormous innovation compared to passing a file back and forth, and we’ve had it long enough to take for granted its many efficiencies. Reality Catches Up to Perception People who have been living inside these tools for a couple of years feel like the ground has shifted completely. Then they look up and see everyone still at their desks doing jobs that should have evaporated, and it seems inexplicable. I’ve been as immersed as anyone, and this project has been a useful correction. The explanation is sitting in the room. There are people in it. If this were a team of three, it would already be mostly automated and I’d be arguing the other side. It isn’t a team of three. So the long tail will expand, and it will expand dramatically. The disposable things will get made, and they’ll be worth making. But there will still be a head to that tail — slow, expensive, collectively negotiated, anything but disposable. A disposable lens works because one person wears it for one day. What stays expensive is whatever a great many people have to see through at the same time.