Something weird’s happening in design.
Not a trend. Not a color palette. Not a new UX “law” being shared on LinkedIn. This one’s bigger.
It starts when instead of sketching a layout or ideating on a design system, you prompt a trained LLM.
A sentence or two. “An app that feels like a Japanese coffee shop on a rainy day.” Boom…design.
Layout, type, color palette. All done.
It's not product design anymore. Not really. You’re vibe designing. And yeah—it looks good. It feels new. It has energy.
But you could be building a product no one can actually use.
Vibe Design ≠ Aesthetic-First Design
Let’s define the thing before it drifts too far.
Vibe design isn’t “designing a mood.” We’ve been doing that forever. Every product guideline that says “we want to feel human, warm, and professional” is a vibe. But currently the vibe is the input, not just the goal.
We’re working with tools: LLMs, AI agents, increasingly design-native platforms, that let us design not from structure, but from feeling. You don’t decide the function/layout and mood separately. You tell the model the mood, and it gives you the layout. The buttons. The type. The transitions. Everything.
This is the design world’s sibling of vibe coding.
Developers figured this out first. You’ve seen it—people writing prompts like “make me a homepage for a pet adoption startup in Tailwind CSS, soft palette, chunky type,” and getting a usable frontend in seconds. That’s vibe coding: generating code from an aesthetic prompt.
Now designers are doing the same. Not just mocking things up quickly, but using AI to create entire product flows from a loose emotional direction.
And that changes everything.
The Tools Are… Kind of Magic
There are tools out now. Some are public, some just barely scraping into private alpha. You feed in a paragraph, maybe a few screenshots or sketches, and out comes a product shell that looks fully real. It feels polished. It feels… designed.
It might be a model trained on Dribbble and Behance spitting out atomic components that match your brand tone. Or maybe a custom LLM suggesting interface tweaks to “align better with the user’s emotional state.”
It’s fast. It’s good. It’s fun.
But here’s the uncomfortable part: it’s designing without understanding. It doesn’t know who your users are. It doesn’t care if your app has complex flows, or if someone tries to sign up from a jailbroken Android in a rural town with 2G signal and a cracked screen while riding a bike (don't do that).
It just knows the vibe.
Accidentally Designing a Fully-Formed Mistake
The scary part is how far you can get.
You can build a fully skinned product without ever asking the important questions. It works beautifully in demos. The transitions are smooth. The typography feels great. The whole thing pulses with energy. The c-suite loves it.
And then a real user tries to onboard.
That soft-glow, full-screen modal you loved? Blocks the keyboard on mobile.
That post-german maximalist text that fades in? Breaks accessibility.
The form with all the personality? Completely collapses when a user pastes in an emoji.
You’ve designed yourself into a corner. Because you started with the vibe, and forgot the system.
Design isn’t just the surface. It’s behavior, edge cases, error states, user fatigue. It’s what happens after someone uses your product for the third time and is just trying to get to the part that matters.
LLMs don’t know that. They’re brilliant at the craft of aesthetics, but they don’t feel pain when an interface fails. You do.
The Edge Case Black Hole
When you vibe design with AI, you’re generating something from a statistical average of what’s looked good before.
But edge cases? They’re anti-average. They’re rare, messy, human moments. They don’t show up in moodboards.
-
A loading spinner when someone has a weak signal.
-
A state where two users edited the same record at once.
-
The edge flow when a user changes their subscription tier halfway through an expired session.
Designers miss those details when they’re not looking for them. Vibe-based systems almost never include them, and users end up paying the price.
It’s not the prompt that gets blamed. It’s the product.
Good Designers Just Got Way More Important
This might sound like a hit piece on AI. It’s not.
AI tools are amazing. They speed up experimentation, make good design more accessible, and help designers move faster.
But they don’t replace good designers. They will replace bad ones.
Because tools don’t think in systems. They don’t understand context, or intent, or the chain reaction one design decision can trigger across an entire product. That’s orchestration—and it still takes a human to do it well.
AI can handle the craft. It can mimic taste. But it can’t feel tension. It can’t hold ambiguity. That’s the part only a designer brings—and that’s the part that matters most.
You have to hold the edge cases in your head. You have to know what this looks like at scale. On old phones. With bad inputs. When someone’s angry, tired, drunk, or distracted. Not pleasing the C-suite with pretty Dribbble designs, but helping the user.
Even with AI in the loop, design isn’t just about getting pixels on the screen. It’s about knowing why they’re there, how they’ll be used, what could go wrong, and what it should feel like when it’s right.
Don’t Let Vibe Design Replace Product Thinking
We’re now in a world where you can build something that feels amazing from a single prompt. And that’s tempting. But don’t confuse the feeling for the function.
A good product has vibe. A great product has resilience.
It handles weird behavior. It guides. It recovers. It works when people are at their worst.
AI can make the vibe easy. But the responsibility of making it usable? That’s still on you.
This isn’t the kind of thing that resolves cleanly.
So What Should Designers Do?
I have no “do this instead” list here. You’re going to use these tools. You should. They’re powerful. But they don’t care about your users. You do.
So keep the LLMs. Get weird.
Don’t forget to test the boring flows.
Don’t forget to stop and ask, “What does this actually need to do?”
