Design Isn't Being Replaced. It's Being Promoted.
Part 1 of 3: The moral case for designers in the age of infinite software
Where we are
The dominant story about AI and design is a replacement story. The tools can generate layouts, write copy, produce components, ship a working interface in an afternoon — so the designer is the next line item. It’s a tidy narrative and it’s wrong, and it’s wrong in a way that matters, because it gets the direction of the change exactly backwards.
When production is expensive, production is the differentiator. You win by being able to build the thing. When production is cheap, everyone can build the thing, and you win on whether the thing is worth using.
That’s where we are. The cost of making has collapsed. The cost of judgment hasn’t moved at all. Which means design — the actual discipline, not the deliverable — just became the scarce input in every product decision that matters.
The math nobody wants to do
Here’s what the next few years look like in raw volume. Small teams shipping what used to take a company. Solo builders shipping what used to take a team. Products spinning up in days, features in hours. The number of interfaces competing for a given person’s attention is going to increase by an order of magnitude, and none of the underlying constraints on that person are going to change at all.
They still have the same twenty-four hours. The same finite patience. The same limited working memory for figuring out where the button went. Human attention is the one input in this system that isn’t getting cheaper, and we’re about to point a firehose at it.
Infinite supply of product, fixed supply of attention. And in that world, every badly-designed thing that ships isn’t just a weak competitor — it’s a tax. It takes minutes from someone who wasn’t going to get them back. Multiply it by the volume that’s coming.
We are about to waste an enormous amount of human life on interfaces that didn’t need to be confusing.
That’s the promotion. Most of us haven’t noticed it happened.
Why this is a moral claim and not a craft complaint
Look at what bad design actually does, with the aesthetics stripped out. It takes time from people. It makes someone feel stupid for not understanding a system that was never explained. It buries the thing they needed under the thing you wanted them to see. It turns a two-minute task into a fifteen-minute one and then does that to a hundred thousand people. Those are harms. Small, distributed, invisible in aggregate, and completely real.
When designers name these harms, it gets heard as aesthetic grievance — wrong typeface, sloppy spacing, taste policing. That’s the dismissal. It doesn’t survive the list above.
We’ve tolerated them because the alternative was expensive. Doing it right meant more time, more people, more money, and there was a defensible tradeoff to make. That excuse is evaporating. When the cost of building the good version approaches the cost of building the sloppy version, shipping the sloppy version stops being a resource constraint and starts being a choice.
And choices are the kind of thing you can be responsible for.
The obligation runs two directions
The first direction is the obvious one. If you have judgment about this and other people don’t, you owe them access to it. Not gatekeeping it, not hoarding it as job security, not treating “taste” as an ineffable gift that justifies your seat at the table. Most of the people shipping now aren’t designers and never will be, and they’re not shipping badly out of malice. They’re shipping badly because nobody ever showed them what good looks like or gave them a way to check.
This is a solvable problem, and the fix is always the same shape. Start by interviewing people on their taste — what they respond to, what they’d reject, and why — because most people have more of it than they can articulate. Turn what you learn into checks a tool can run, and run those checks while the person is working, so the mistake gets caught the moment it’s made, not in a design review three weeks later that nobody scheduled. Then keep learning: every choice they make, and especially every correction, sharpens the standard — and the standard guides them back, proposing their own best instincts to them until the taste they couldn’t articulate becomes taste they can use. That’s the thing I’ve been building, and it’s why this series exists.
The second direction is the one that actually keeps me up. We don’t just owe it to the builders. We owe it to everyone downstream who has to use what gets built, who never consented to being the test surface for someone’s fast shipping, who just wanted to pay an invoice or book a flight or find out when the thing arrives.
Those people have no representation in the room. They’re not on the roadmap review. They can’t push back on the sprint. The only advocate they’ve ever had is the person in the process whose job was to ask whether this makes sense to someone who doesn’t already know how it works.
The obvious objection: aren’t the models developing taste too? They are — they can already evaluate work against a standard, and they’re getting better at it fast. But evaluating against a standard and setting the standard are different jobs. Someone still decides what good means here, for these people, at this moment — and owns the consequences when the call is wrong. Judgment without accountability is just output. That’s the part that doesn’t automate, and it’s the part this series is about.
That role doesn’t disappear when the code writes itself, because it requires having modeled a stranger’s confusion and cared about it.
What we’re actually for now
The job was never making the artifact. That was just the medium available for expressing judgment. When you take away the labor of production, what’s left is the judgment — and it turns out the judgment was always the product.
So the work becomes: deciding what deserves to exist. Knowing which of six plausible flows survives contact with a tired person at eleven at night. Noticing that the interface technically works and is still going to make someone feel like an idiot. Holding a line on quality when the pressure is entirely toward volume, because volume is now free and quality still isn’t.
That’s not a smaller job than the one we had. It’s a much larger one, and it comes with actual stakes attached. Every product that ships is a claim on someone’s finite time. We’re the ones who decide whether that claim was worth making.
More things are going to get built in the next five years than in the previous fifty. Somebody has to care whether they’re any good.
It should be us. It has to be us.
Next in the series — Part 2: making good design accessible through Raven — how a taste engine turns judgment from something you have into something you can hand to anyone who ships. Part 3: the selfish reason — building agents for the parts of design I don’t feel like doing that day.