WeChat is quietly testing an AI assistant called Xiaowei. Swipe right from the home screen and it’s there. Tell it “get me a milk tea, light sugar, light ice” and it goes off, opens the right mini-program, places the order, pays — you just tap to confirm. Say a sentence and it’ll build you a small app in a few minutes: an expense tracker, a mood journal, ready to use on the spot.
The headlines all landed on the same line: WeChat is finally becoming the super-app. Every company now works for it, every app eventually decays into an API, and WeChat becomes the one and only front end.
That’s true, but it’s the surface. I don’t want to argue about whether this is a smart move — right now it’s rough, and it falls apart the second it hits one of those travel apps where pop-ups breed pop-ups. What I want to do is follow the direction it’s pointing a few steps further, see what’s actually at the end of that road, and what it means for a designer.
You won’t open the food app. You’ll just say “I’m hungry.”
Push this to its conclusion and the picture looks roughly like this. One day you stop poking at icons one by one. You talk to WeChat, and it calls everything for you. The food app, the shopping app, the travel app — the hundreds of apps on your phone all shrink behind it and become “data sources” it reaches into. You don’t “open the food app” anymore. You just say “I’m hungry.” Every company except WeChat retreats backstage and offers nothing but an interface. WeChat, as the only front end, becomes the single door to every service.
This isn’t WeChat’s script alone. Every assistant out there — Doubao, Qwen, all of them — is fighting for the same thing: the instant a thought forms in your head, satisfy it right there. WeChat just happens to hold the best hand. It has payments, it has your identity, and it has a million ready-made mini-programs sitting there to call.
Say this future actually arrives. What does it mean for a designer? That’s the part I actually want to talk about.
UI design is about to become a heritage craft
To see what’s happening you have to say the quiet part about UI out loud: at its core, UI design is a translation layer between humans and machines. It’s an API protocol between the two.
Machines don’t understand human speech, and humans don’t speak machine. Something has to sit in between and turn your intent into instructions the machine can run, then turn the machine’s state into a picture you can read. A UI designer’s whole job, stripped down, is making that translation smoother — taking some friction out of the handoff between person and machine.
But AI just finished that translation for you. It understands your mumbled “I’m hungry” and it can drive the backend interfaces. The UI layer in the middle no longer has a reason to exist. You don’t hunt across a row of buttons anymore. You say a sentence and the thing is done. There’s a line in The Three-Body Problem where the aliens seize the particle colliders and so “physics no longer exists.” When AI takes over the human-machine interface, UI design no longer exists either.
Plenty of designers are still saying AI will learn to make UI and steal their jobs. The truth may be that AI doesn’t take the bowl — it smashes it. At the root, UI design loses its reason to be. Everything we treat as professional craft today — nudging pixels, dialing in spacing, usability, optimizing a conversion funnel layer by layer — becomes worthless fast. Because the whole act of UI design is on its way out.
This already happened, a hundred years ago
None of this is new. When the camera showed up a century ago, painters lived through exactly this.
Before it, a painter’s most valuable skill was making it look real — light, perspective, rendering a fold of cloth so it read as silk. That kind of realism took a lifetime to train. Then the camera arrived and “painting it accurately” was worthless overnight. One press of the shutter beat a month of your brushwork, and it was more accurate.
But what the camera killed wasn’t painting. It was the skill of making it look real. The craftsmen who lived only off that skill got replaced. The people who’d always been using the image to tell a story, carry an emotion, work out what they were actually trying to say — they got more valuable. For them, realism was a tool, not the whole game. The camera actually kicked painting forward: Impressionism, Fauvism, Cubism all detonated after it. Once the grunt work of recording reality went to the machine, what was left for humans was making the things reality doesn’t contain.
UI design stands at that same fork today. What AI kills is the part that only aligns boxes and assembles screens to spec. That skill loses its value the moment a new tool shows up — no different from the darkroom technicians who only knew how to develop film.
90%+ of designers become useless
The ones getting washed out aren’t a few individuals. It’s the vast majority.
A lot of designers carry a kind of competence illusion. They run the software fluently, they’ve internalized the methodology, they know how to optimize an experience against a metric — and all of it is depreciating fast. Not because AI does these things better than you. Because the future simply won’t need them done at all.
What survives, and keeps getting more valuable, is creation — making the things AI has never seen, the things that aren’t in its training data.
Why that and nothing else? Because AI grows up eating what humans already made. At its core it’s the strongest stitching machine ever built — fast, convincing seams. But at least today, what it’s good at is recombining things it has seen. The genuinely unseen still needs a human to make it first and feed it in. The person who can produce “training data AI has never seen” is the one position AI can’t replace. And here’s the sharper part: the new thing a human makes immediately becomes the next batch of food fed back to it. The human stays half a step ahead, forever; the AI is always behind, chasing. That half-step is the only ground a designer has left to stand on.
A survival guide for designers
Let me stick my neck out with a few calls about where this profession is headed. They’ll need time to prove out.
First: climb out of the screen and back into the physical world.
If UI is the API between humans and machines, follow the thought down one level — industrial design is the API between humans and the physical world. A chair, a piece of furniture, a car. AI can’t take this layer over anytime soon. Not because it can’t learn industrial design, but because the user of industrial design is a real human body. UI, by comparison, was never a real human need. Your hand still has to close around a real object, your body still lives in the physical world. Short of uploading ourselves into the cloud, that need doesn’t go away.
The physical buttons coming back to cars are a tell. For a while everything got buried in the touchscreen because that looked advanced. But driving puts you in a passive, receiving state — you can’t take your eyes off the road to hunt through a screen just to change the temperature. A physical knob your hand finds without looking is the right answer. Once AI flattens the digital interface layer, how humans deal with the physical world only matters more.
Second: turn yourself back from a problem-solver into a problem-asker.
This one’s the easiest to miss. AI is the most powerful problem-solving machine humans have ever built: hand it a question and it answers fast and well. But it has a flaw — it doesn’t ask. You say you want A and it’ll hand you the best possible A. It won’t stop and push back: “Are you sure A is what you want? Could the real problem be B?” It has no desire, no itch to know.
And the designer happens to hold the one thing asking demands: a trained eye. Put the same object in front of two people and the layman sees roughly what’s there, while a good designer sees a few hundred small things that are off, things nobody else catches — where a step took one detour too many, where something made the gut clench. That sensitivity to what’s off is where a good question starts. AI can solve anything you hand it, but it won’t spot what’s wrong for you, and it certainly won’t judge which problem in the pile is the one actually worth solving.
So what’s genuinely scarce in the future isn’t answers. It’s questions. The people who mattered most in history were rarely the ones with their heads down solving. They were the ones who asked the right thing first.
When answers go infinitely cheap, when anyone can summon an army of AI to solve their problems, asking a high-value question gets even more precious. A good question is the edge of one person’s imagination — and that edge is something AI can’t draw. Only a human can.
Look closely and both routes are the same sentence: stop being the one with your head down, executing. Move to an API layer AI can’t touch, and train an eye AI doesn’t have — the eye that finds the problem. It comes down to going back into creation and out of the grunt work.
But does being able to create actually guarantee a future?
That all sounds inspiring: designers go back to creating, make the things AI has never seen. Sit with it a beat longer, though, and I feel a little bleak.
In this loop, what role do humans actually play? We’re the producers of training data.
Everyone says AI is going to eat the world, but what does it actually need? On the surface, energy and chips. Except energy isn’t scarce and chips can be mass-produced — both keep getting cheaper. The thing that’s truly scarce, can’t be mass-produced, and only humans can keep generating without stop, is the data — fresh human creation that AI hasn’t seen yet. To the AI, that data is the ultimate battery.
This is The Matrix, almost exactly. Except in the film humans were the biological cells powering the machines, and here, now, we’ve become the cells feeding the AI its “data.” We think we’re creating, self-actualizing, outrunning the machine. Look at it another way, and maybe we’re just feeding it, endlessly.
Which makes me think of something. Billions of years ago, the earliest anaerobic bacteria pumped out enormous amounts of oxygen and, bit by bit, assembled an oxygen atmosphere — which paved the road for the oxygen-breathing life that came after. Then the anaerobes exited and the oxygen-breathers took the stage. They toiled through an entire existence, and in the end all they’d done was build the conditions for the next form of life.
So all of human civilization, all of human intelligence — is it maybe just the oxygen we’re laying down for AI, the next form of life? We cast ourselves as the protagonist, but underneath, are we just that batch of anaerobic bacteria? Only this time what we produce isn’t oxygen. It’s data.
At least today, when we tell designers to go back to creation, this question is worth keeping somewhere in the back of your mind. Is it a way out — or just the new workstation the bigger game has assigned to the designer?