The handoff that made me snap
I still remember the exact screen. It was a dealer inventory dashboard I had spent three weeks designing for Valmont. I had pixel-pushed every state: loading, empty, error, edge cases with 200+ line items. The Figma file was pristine. I presented it, got approval, and handed it off with a detailed annotation document.
Two months later I saw the live version. The spacing was off. The empty state I had carefully designed was replaced with 'No data found' in 12px grey text. A button I had placed based on user research was moved because the developer 'thought it looked better on the right.' I was furious. And then I was embarrassed — because I had no power to fix it. I could only file a Jira ticket and wait.
That was the moment I decided: I am going to learn how to build what I design. Not become a full-time engineer. Just enough to ship my own screens end to end. No handoff. No waiting. No compromises I did not agree to.
The first thing I built was embarrassing
I started where most designers start: with zero coding knowledge beyond basic HTML and CSS from design school. My first attempt was a simple landing page for a side project. I opened VS Code and stared at the screen for 20 minutes because I did not know what `npx create-next-app` meant. I watched YouTube tutorials that assumed I knew what a terminal was. I copied code from Stack Overflow and broke things in ways I could not explain.
The page took me two weeks. It had three sections and a contact form that did not actually send email. But when I opened it in a browser and saw the exact design I had imagined — my padding, my type scale, my colour choices — all working in a real browser, not a mockup... that feeling was different. It was not a picture of a product. It was a product.
The quality was terrible by engineering standards. But it was mine. And more importantly, nobody had misunderstood my intent between Figma and production.
What AI actually changed for me
Here is the honest version. If I had tried this in 2022, I would have given up. Learning to code from scratch while working full-time as a designer is unrealistic for most people. What changed everything was AI coding assistants.
I use Claude for the architectural thinking — explaining what I want to build, asking it to suggest a component structure, getting it to walk me through the data flow. I use OpenCode for the actual implementation inside my editor. The workflow looks like this: I design the screen in Figma, export the key patterns, describe the interaction in plain English, and let the AI generate the initial code. Then I review, tweak, and iterate — just like I would with a junior developer.
The AI is wrong maybe 40% of the time on the first pass. It hallucinates imports. It uses deprecated patterns. It over-engineers simple things. But it is fast. I can go from idea to working prototype in hours, not weeks. And because I understand the design intent — because I am the designer — I catch the mistakes quickly. The AI writes the code. I provide the judgment.
What got easier (and what got harder)
The obvious win: speed. A feature that used to take two sprints — design, handoff, dev, QA, revision, deploy — now takes me a few days end to end. No meetings with engineers to explain spacing. No tickets sitting in a backlog. I see a problem, I fix it, I ship it.
The unexpected win: my design quality improved. When you build what you design, you quickly learn which interactions are easy to implement and which are a nightmare. You stop designing fancy animations for components that would require 200 lines of custom JavaScript. You start respecting platform conventions because you feel the pain of fighting them. You become a more practical designer.
But some things got harder. Context switching between design brain and development brain is exhausting. Some days I spend six hours debugging a state management issue and make zero design progress. The tools are improving — but there is still a gap between 'the AI generated it' and 'it is production-ready.' I still rely on actual engineers for complex backend work, authentication, and deployment. I am not pretending to be a full-stack developer.
Who should try this (and who should not)
If you are a designer who is curious about code, start small. Build a single page. Ship it. See how it feels. You do not need to quit Figma or call yourself a 'design engineer.' Just add one tool to your toolbox.
If you are a designer who has zero interest in code, that is fine too. The industry does not need every designer to code. What the industry needs is designers who understand enough about implementation to design with real constraints in mind — and developers who respect design intent. The handoff problem is not going to be solved by everyone learning to code. It is going to be solved by better collaboration, better tools, and AI bridging the gap.
For me, building what I design has been the most rewarding shift in my career. Not because I love writing code — I do not. I love seeing my designs work in the real world, with real users, without anyone 'simplifying' them along the way. That is what got me into design in the first place.