Designers Are Building Real Things Now. Here's My Honest Take on How.

Two years ago, if you told me I'd be shipping a live mobile app — with authentication, in-app purchases, and real users — without a development partner, I would have laughed. Not skeptically. Actually laughed.

I'm a designer. Seventeen years in. I know Figma like a second language. I can prototype in Framer until something feels alive. My craft is not the question.

But building? Shipping? That always required someone else. You handed off the file. You waited. You hoped the sprint planning meeting didn't kill it. You got back something that was close but not quite right, and you smiled and said it was fine.

That bottleneck is breaking. And I have receipts.

What actually changed

I built my own app. A faith-based mobile product I'd been sitting on for over a year. Not a prototype. Not a Figma file. A real, live thing with a backend, subscriptions, and users who paid for it.

The stack: Replit for development and deployment. Claude as the AI brain doing the actual building. Me as the designer, the product thinker, the person with taste calling the shots on every decision.

I'm not a developer. I still can't write a clean React component from scratch without help. But I understand how things are supposed to feel. I know when spacing is off. I know when a flow has too much friction. I know what the product is supposed to do and what it should never do. That knowledge — the designer's knowledge — turns out to be the most valuable thing you can bring to this workflow.

Claude Code and Replit are the combination I'd point any designer toward. Claude Code operates like a tireless engineering partner. You describe what you're building in plain language, it reads your files, writes the code, and builds incrementally. Replit gives you a live environment without touching a terminal. You can see your app running. You can share a link. You can iterate in real time.

The tools are not magic. They require clear thinking, tight prompts, and knowing how to review what gets built. But they put the power to ship directly in the hands of the person with the vision — which, in most product conversations, is the designer.

But I'm not throwing away my tools

Here's where I push back on the take I keep seeing.

Some people are treating vibe coding like it replaces the craft. Like Figma is legacy software and Framer is just a stepping stone to a real deployment. That's wrong, and I think it misunderstands what design tools are actually for.

Figma is not a handoff tool. Figma is where I think. It's where I make decisions before anything gets built. Colors, spacing, component behavior, layout logic — all of that happens in Figma first. When I sit down with Claude to start building, I have a Figma file open on the other monitor. I'm not describing vibes. I'm describing a designed thing.

Framer is still one of the most powerful prototyping environments available. When I need to communicate motion, interaction, or the feel of a real UI, Framer is the place. AI coding agents are not good at capturing nuanced animation intent from a text description. A Framer prototype is. You show the agent the behavior. You describe it specifically. That combination gets you closer to the real thing faster.

The designers winning with AI tooling are not the ones who abandoned their craft. They're the ones who use their craft to direct the machine. Bad prompts produce mediocre output. Knowing your fundamentals — typography, visual hierarchy, component architecture, system thinking — is what separates a designer building a real product from someone generating slop and calling it shipped.

The actual workflow

Design first in Figma. Real decisions, real colors, real component logic. Not a wireframe you plan to figure out later. The more specific the design, the better the output from the agent.

Use Claude Code or Replit Agent to build section by section. Not the whole product in one shot. One component. One screen. One interaction. Review it against the design. Adjust. Move to the next thing. Tight loops beat big batches.

Prototype edge cases and motion in Framer when you need to communicate something the agent can't infer from text. Show, then describe.

Understand what the code is doing without needing to write it. Know that there's a components folder. Know that the app has routes. Know that a button is a component with states. This mental model costs you nothing to build and dramatically improves the instructions you give.

Know when to bring in a developer. Complex animations, scaling infrastructure, deep integrations — that's still human territory. The goal isn't to replace development. The goal is to get your idea into the real world before momentum dies.

What this actually means

Developers have always been able to go from idea to product. Designers have had to pitch, wait, negotiate, and compromise. The ability to ship your own idea — test it with real users, charge real money, learn from real feedback — is a different relationship with your work entirely.

That doesn't happen by abandoning the craft. It happens by adding to it.

Figma. Framer. Claude. Replit. In that order, more or less.

The best designers right now are not choosing between craft and code. They're dangerous in both directions.