Last week, I turned a Figma mock-up into a working prototype in 12 minutes. Not a clickable design prototype. Not a “looks like it works” animation.
It was functional. Testable. Real. Users could break it, navigate it, experience it. No handoffs. No compromises. No “that’s not technically feasible” conversations.
This is vibe coding, and it’s about to change everything about how you work.
Enter Vibe Coding (Yes, That's Really What It's Called)
Coined by Andrej Karpathy in early 2025, vibe coding is beautifully simple: describe what you want in plain English, get working code back. No syntax wrestling. No three-day battles with CSS Grid.
It’s less “Create a 1200px container with 24px padding, #FF6B6B background, and ensure the CTA button has a 4px border radius with hover states.” And more “Make this feel like a Berlin rave — energetic, slightly chaotic, but ultimately welcoming.”
The AI gets the vibe. You get the result.
Why This Changes Everything Right Now
Your clients are already seeing AI-generated work. Your competitors are already using these tools. Your colleagues are definitely experimenting with them. You can either lead this transition or get dragged through it.
What most designers get wrong is they think this is about replacing creativity. It’s not. It’s about replacing the tedious parts of design so you can focus on the strategic thinking that actually matters.
Here’s what’s already happening:
- 95% of code in top Y Combinator startups is now AI-generated
- Design-to-prototype cycles have collapsed from weeks to hours
- Non-technical team members are building functional applications
- Client expectations for speed and iteration have permanently shifted
This isn't coming. It's here.
You don't have to speak Parseltongue to understand Python
Your New Superpowers
Need For Speed
Turn Figma designs into live demos instantly. Test multiple concepts with real users while your coffee’s still warm. A/B test different aesthetic approaches without begging developers for “just five minutes” of their time.
Your Role Just Got More Interesting
You’re no longer the person who makes things look pretty. You’re the person who makes things work beautifully. Less time pushing pixels, more time solving actual problems.
True Collaboration
You no longer need to speak Parseltongue to understand Python. Cross-functional teams can contribute through natural language. PMs, writers, developers, designers — everyone speaks the same language now. No more translation layers. No more momentum-killing conversations.
Week 1 Starter Guide
Choose one tool. I recommend Lovable because it won’t make you question your career choices. Create an account, understand your limits, watch exactly one tutorial. Don’t fall down the YouTube rabbit hole of “Advanced AI Prompting Techniques.” You’re not ready, and you don’t need to be. Here's some options...
If you live if Figma:
- Anima — Direct Figma to working code (actually works)
- Lovable — Design-focused with UX best practices baked in
- Figma Make — AI-powered plugin for production-ready code
If you like starting fresh:
- Bolt.new — Zero to app with minimal faff
- Replit — Full dev environment with AI assistance
- Vercel — Git-to-live-site in seconds
Pick one. Try it for a week. Most free accounts limit you to a few prompts daily — treat this as efficiency training.
Step 2: Start Embarrassingly Simple
Take your most boring current project. Not the rebrand that’ll define your career. The simple stuff. A contact form. A basic landing page. Something that won’t matter if it looks like a caffeinated intern designed it.Write one sentence: “A clean contact form that feels trustworthy.” Hit enter. See what happens.
Step 3: Make it Real
Use actual copy, not lorem ipsum. Add your real brand colours. Show it to a colleague for exactly 30 seconds of feedback. If they say “that’s not terrible,” you’re winning.
Step 4 & Beyond: Make it Dangerous
Now you can start being ambitious. Take a real project. Something that matters. Use what you’ve learned to create prototypes that actually matter. Test them with real users. Break things. Fix them. Repeat.
Perfect for:
- Rapid prototyping and concept validation
- Interactive prototypes for user research
- Stakeholder demos where static mock-ups make you look dated
- A/B testing different emotional approaches quickly
- Landing page concepts you need to test without begging developers
Don't use for:
- Complex user flows with seventeen different states
- Anything involving actual user data
- A full web project (not yet!)
- Projects where accessibility is critical (AI is terrible at this)
- Anything that actually matters without human oversight
Please Vibe Responsibly
- Start Simple. One clear sentence beats a novel-length prompt
- Import from Figma. When possible (it genuinely helps)
- Break Down Complexity. AI isn’t magic
- Always Review Output. Trust but verify
- Combine with Research. AI doesn’t know your users
- Keep Critical Thinking Switched On. Someone has to
Pro tip: Use ChatGPT to refine your prompts before feeding them to your development platform. Meta? Yes. Effective? Absolutely.
The Human Edge
The robot overlords haven’t completely taken over yet. Here’s what you need to protect:
Strategic Thinking
Understanding what users actually need based on evidence, not assumptions. Creative vision that goes beyond “make it pop.” Problem prioritisation based on real business goals.
Design Judgement
Nuanced UX decisions. Brand consistency that doesn’t follow every trend. Accessibility and inclusive design (AI still thinks everyone has perfect vision and motor skills).
Technical Oversight
Architecture decisions for complex applications. Security reviews (AI loves creating vulnerabilities). Debugging when things inevitably break.
Legal Stuff. And Other Serious Things
This isn’t all sunshine and rapid prototypes. Here’s what you’re signing up for:
The Legal Minefield
IP theft risks. Murky ethics around training data. Zero protection for your original work.
Professional Limitations
AI won’t challenge your dodgy design decisions. No creative critique. No contextual understanding of actual user needs. Basically a very clever parrot.
Quality Control
Surface-level solutions without proper research. Generic outputs making everything look the same. Technical debt that’ll bite you later.
Maintenance
Good luck updating code you didn’t write. Debugging AI-generated code is like solving puzzles blindfolded. Security vulnerabilities hiding in plain sight.
Start Building The Future
Vibe coding won’t solve your fundamental design problems. Bad content is still bad content, whether AI-generated or not.
But it will let you test ideas at the speed of thought. Validate concepts with real users instead of defending wireframes in conference rooms. Turn creative vision into working reality without waiting for development resources.
The future is collaborative — humans driving creative vision while AI handles the grunt work. Focus on users, not grunt work. Let machines handle the details. Human judgment trumps all.
The real question isn’t whether you should use AI tools. It’s whether you’ll lead this change or get left behind explaining why your mock-ups don’t match the final product.
Stop overthinking it. Pick a tool. Build something embarrassingly simple. Show it to someone. Repeat until it doesn’t suck.
Welcome to the future. It’s messier than the tutorials suggest, but infinitely more interesting than pushing pixels around Figma for the rest of your career.
Ordered list
