Vibe Design Is Here. Your Figma Subscription Just Got Nervous.
Remember when Andrej Karpathy coined “vibe coding” and half the internet lost its mind? Developers split into two camps overnight: those who saw the future and those who saw the apocalypse. Nobody stayed neutral.
Well, it took about five months for someone to apply the same logic to design. And this time, it’s not a tweet that started it — it’s an actual product.
Claude Design launched two days ago, and a new term is already making the rounds: vibe design. As in, you describe what you want, tweak some sliders, and ship a prototype without ever opening Figma. No static mockups. No artboard archaeology. Just vibes.
If that sounds familiar, it should. We’ve been living through the vibe coding revolution for months. This is the same movie, different screen.
What “Vibe Design” Actually Means
Let’s get specific before this turns into another buzzword nobody can define.
Julian Oczkowski, a designer who spent significant time testing Claude Design, defined vibe design around three ideas:
You talk, it builds. Instead of dragging rectangles around a canvas, you describe your intent in plain English. “I need an onboarding flow for a meditation app, dark mode, minimal, no more than three steps.” Claude generates a first version. You have a conversation about what to change. It rebuilds. You adjust a slider. It rebuilds again.
There’s no file. This is the radical part. Traditional design produces a .fig or .sketch file that a developer then rebuilds in code. Vibe design skips that step entirely. The design IS the prototype. The prototype IS the handoff. One user testing Claude Design reported that “no Figma file was produced at any point.” The output went straight from conversation to interactive HTML to a Claude Code implementation bundle.
Parameters replace pixels. Instead of nudging elements 2px to the left for the fourteenth time, you adjust high-level controls. Typography scale. Color temperature. Spacing rhythm. The AI handles the pixel math. You handle the intent.
Think of it like the difference between writing HTML by hand and using a CSS framework. You’re not abandoning control — you’re operating at a higher level of abstraction.
The Taste Problem (Or: Why Vibes Aren’t Enough)
Here’s where things get honest.
One of the sharpest analyses came from a Substack writer who tested Claude Design on a real project. The AI extracted the brand colors correctly. It understood the typography. It identified the right component patterns. And then — and this is the part that matters — it composed all of those correct ingredients into a design that a trained eye would reject in two seconds.
It’s like giving someone all the right spices and watching them make a terrible curry. The ingredients were perfect. The cooking was off.
This is what designers call taste, and it’s the thing AI consistently stumbles on. Taste isn’t knowing that your primary color is #F7931A. It’s knowing that using it on 40% of the page makes the whole thing look like a traffic warning. Taste is the funeral home not using the same playful rounded corners as a children’s toy store, even though both technically pass accessibility checks.
A Hacker News commenter put it perfectly: “You’ll get a competent UI with little effort, but nothing truly unique or mind-blowing.” And they’re right. Claude Design produces output that sits comfortably in the 60th percentile. Professional enough to ship internally. Generic enough that your SaaS dashboard will look exactly like everyone else’s SaaS dashboard.
For a lot of use cases, that’s genuinely fine. For anything that needs to stand out? You still need a human who has opinions.
The 60% That’s Free (And the 40% That’s Expensive)
Here’s a framework that keeps showing up in designer discussions: the 60/40 split.
AI gets you to 60% fast. Scary fast. Wireframes in seconds. Brand-matched prototypes in minutes. Landing pages before your second coffee. That 60% used to take days. Now it takes a conversation.
But the remaining 40% — the part where a designer decides that this particular flow needs an unexpected moment of delight, or that the onboarding should deliberately slow down at step two because user research showed people feel overwhelmed — that part hasn’t gotten any cheaper. If anything, it’s gotten more valuable because everyone’s 60% now looks the same.
It’s Jevons Paradox applied to design. When something gets cheaper, demand doesn’t decrease. It shifts. What used to be valuable (execution speed) becomes commodified. What was always valuable but hard to measure (judgment, taste, strategic thinking) becomes the whole game.
The data backs this up. Design skills — not Figma skills, not wireframing skills, but actual design thinking — are now the number one most in-demand capability in AI-related job postings. Ahead of programming. Ahead of cloud infrastructure. Companies building AI products are desperate for people who understand how humans interact with interfaces, because the AI can build the interface but it can’t tell you if it should exist.
Who Wins, Who Loses, Who Adapts
Let’s be blunt.
Winners: Senior designers who were already thinking strategically. They just got the biggest productivity boost in the history of their craft. One senior designer with Claude Design can now prototype more in a week than a five-person team could six months ago. The time they used to spend in Figma moving boxes around now goes toward research, strategy, and taste. That’s a trade-up.
Winners (unexpected): Developers, product managers, founders — anyone who had ideas but couldn’t visualize them. Vibe design is the design equivalent of what vibe coding did for non-programmers. Your PM can now walk into a meeting with a clickable prototype instead of a napkin sketch. That’s genuinely powerful.
Losers (for now): Mid-level production designers whose value was primarily “I turn wireframes into polished Figma files.” That workflow is being compressed. Hard. The PCWorld reviewer burned through 80% of a Claude Pro subscription in 30 minutes, so the economics aren’t quite there yet — but they will be.
Adapters: Everyone else. If you’re a designer reading this and feeling nervous, the move isn’t to fight the tool. It’s to become the person who tells the tool what to build. Learn to evaluate AI output critically. Develop your taste muscle. Get better at the things AI can’t do: user research, stakeholder management, strategic disagreement, and knowing when the technically correct answer is the wrong one.
The Figma Question
People keep asking: is Claude Design a Figma killer?
Short answer: no. Longer answer: it’s a Figma-for-certain-workflows killer.
Figma is a multiplayer design tool with deep collaboration features, design systems, developer handoff specs, and years of team workflow built around it. Claude Design is a single-player conversational tool that burns through tokens fast and doesn’t support co-editing. They’re solving different problems.
But here’s the uncomfortable truth: a huge percentage of Figma usage was never really “design.” It was asset resizing, deck-making, and wireframe iteration that could absolutely be replaced by a good conversation with an AI. Figma knows this. Their stock didn’t drop 7% for nothing.
The real competition isn’t Claude Design vs. Figma. It’s Claude Design plus Figma vs. doing things the old way. The designers who’ll thrive are the ones using both — vibing through the exploration phase with Claude Design, then bringing the best direction into Figma for the precision work that still needs human hands.
Where This Goes Next
If vibe coding taught us anything, it’s that the first version of these tools is never the last. Vibe coding started as a novelty and within months became a legitimate development workflow. Vibe design is on the same trajectory.
Within six months, expect direct manipulation tools to integrate with parametric AI design. Within twelve months, mobile design workflows. Within two years, the entire concept of a “design file” might feel as quaint as a floppy disk.
But taste? Taste isn’t going anywhere. The slot machine might generate a thousand options. Somebody still has to know which one to bet on.
Related Reading
Share this post
Comments
Leave a comment
NativeFirst Team
EditorialThe NativeFirst team — engineers and designers building native Apple apps and writing the courses we wish we had when we started.