I Built My First Website in 1996. Design Has Died Three Times Since.
I built my first website in 1996.
Tables for layout. Fonts that only worked if the visitor happened to have them installed. A loading bar you watched while the modem screamed.
Design has died three times
Or so they keep telling me.
Print to web. Web to mobile. Mobile to AI. Every few years someone announces the end of design, and the obituary is always written with total confidence.
Flash would replace us. Then templates would. Then no-code would. Now AI will. Each wave arrived as a tool that made production easier, and each time the easier the tool got, the louder the panic got. I've now watched this exact movie three times. The poster changes. The plot doesn't.
The tool was never the job
Here's what thirty years actually taught me.
The tool was never the job. The job is deciding what should exist and what shouldn't. Knowing which problem is worth solving and which one is a distraction dressed up as a feature. Knowing what to leave out, which is the decision nobody thanks you for and everybody benefits from.
Tables, Flash, Figma, a prompt. All of it is just notation. It's where the work gets written down. It was never where the work got decided. I changed notation four times across one career and the part that mattered never moved.
AI can generate a hundred screens. It can't tell you which one deserves to exist.
This is the thing the panic keeps missing.
AI can generate a hundred screens in a minute. Good ones. Plausible, polished, professional ones. What it can't do is tell you which of the hundred deserves to be built for this product, this audience, this problem. It answers the brief beautifully and never thinks to ask whether the brief was right.
That decision has never lived in the software. It lived in the person who'd seen enough to know that the obvious answer was a trap, or that the feature everyone was asking for would quietly make the product worse. AI removed the bottleneck between an idea and a shippable thing. I built and shipped a working app in a weekend to prove it to myself. But it didn't remove the need for the idea to be good. If anything it raised the price of a bad one, because now you can build the wrong thing faster than ever.
The craft was never the tool
So here's the reframe that survives every one of these transitions.
The craft was never the tool. The tool was where we hid the craft.
For thirty years we got to point at the hard, visible, learnable thing, the software, and call that the skill. Master the pen tool. Master the grid. Master the framework. The mastery was real, but it was always standing in front of the actual work, which was judgment. The tool was a hiding place. AI just knocked the wall down and exposed what was always behind it.
That's why AI output has no taste no matter how capable the model gets. Taste isn't in the rendering. It's in the choosing, and choosing is exactly the part that never lived in the software in the first place.
The wrong layer
Every designer panicking about AI right now is worried about the wrong layer.
They're watching the execution layer get automated and reading it as the end, because for years the execution layer was where they kept their identity. But the layer that matters has always been the one above it. The deciding. The judgment about what should exist. That layer hasn't been touched, and the tools that automate everything beneath it only make it more valuable, because now it's the only scarce thing left.
It's the same reason depth is the only moat left in software: if a thing can be regenerated in a weekend, it was never the defensible part. Judgment can't be regenerated, because it isn't output. It's the accumulated record of a few thousand decisions about what was worth making.
Design didn't die three times. The tool died three times, and we kept mistaking it for the job. I'm not worried about the fourth. I've just stopped hiding the craft inside the software, because this time the software finally can't hold it.
---
