Design to code without handoff

The shift

Instead of Figma as the final artifact, I treat it as the thinking tool. The actual output is a React component. A live, interactive thing that works in the product's real context.

This means learning enough code to close the gap. Not becoming an engineer. Just enough to own the implementation decision from intent to browser.

Tools matter here. Claude Code lets me iterate fast. Write a component, test it, refine it, ship it. No context switching. No waiting. No translation layer where intent gets lost.

What changed

At CLARK, this approach shifted how we worked on the New Onboarding Journey. I could design a flow, prototype it as real components, and hand over something that already lived in the product's codebase. Not a screenshot. Not a spec. A working thing.

The team could focus on integration instead of interpretation. We hit 36% conversion uplift on that project. Part of that was design rigor. Part was speed. Part was the fact that what was built actually matched what was intended.

The tradeoff

This doesn't scale infinitely. I'm one person designing one flow. If you have ten designers, this model breaks. You need different tools, standards, documentation. That's not what this is.

It also requires a specific kind of permission. Not every company lets designers touch code. Not every team structure supports it. You need trust that the designer knows what they're doing in both directions.

But when it works, it removes a layer of noise. You go from sketch to production artifact in one continuity. No lost conversations. No reinterpretation. Just design and implementation as the same thing.

Why now

Design systems matured. Component libraries exist. The tools got better. Claude Code specifically lets you work at speed without needing to be a professional engineer. You can prototype, test, deploy, all within your thinking loop.

The designer's job becomes: understand the product deeply enough to own both the experience and the code. That's a bigger scope than before. But it's also clearer. You're not managing interfaces between disciplines. You're managing one thing: does this work?

The honest part

This is a differentiator because it's rare. Not because it's universally better. Some of my best work has been pure design collaboration with amazing engineers. The handoff, when done well, produces better results because you have two kinds of expertise.

But for speed, for coherence, for design intent to survive contact with the product, this approach has value.

The question isn't "should all designers code?" The question is "who benefits from closing that gap?" And if you're the kind of designer who thinks in systems and constraints, who gets frustrated by lost intent, who wants to own the full loop: maybe it's worth it.