I Couldn't Code on Friday. I Shipped an App on Sunday.
After 20 years of designing apps for other people, I finally built my own. Just me and Claude Code.
It's called Paak. It helps you pack and prepare smarter for trips. And it actually works. Not a prototype. Not a Figma demo. A real app. Live.
The thing nobody tells you about being a designer is how frustrating it is to always have the vision but never be the one who ships it. That frustration is now optional.
What changed
For two decades, my workflow ended at handoff. I'd design the screens, document the interactions, spec the components, then hand a Figma file to an engineer and wait. Sometimes what came back matched the design. Sometimes it didn't. Either way, I wasn't the one making it real.
AI coding tools changed that equation overnight. Not gradually. Not incrementally. Overnight.
I sat down on a Friday evening with nothing but a Figma design and a Claude Code session. By Sunday night, I had a working app deployed to production.
How the process actually worked
I didn't learn to code. That's the important distinction. I learned to direct code. The difference is like the gap between being a carpenter and being an architect. Both understand buildings, but the work is fundamentally different.
My process was conversational. I'd describe what I wanted in design language ("this component needs to slide in from the right with a 300ms ease-out") and Claude Code would write the implementation. When something didn't look right, I'd describe the gap between what I saw and what I expected, the same way I'd give feedback to a developer.
The mental model that helped most was treating Claude Code like a very fast, very literal junior engineer. It does exactly what you ask. The skill is knowing what to ask for.
What designers should know before trying this
First, your design skills transfer directly. Understanding layout, spacing, hierarchy, and interaction patterns is the hard part of building interfaces. The same skills that make business app design valuable are the ones that translate best to AI-assisted development. The code is just the notation.
Second, you'll hit walls. I spent three hours on a state management issue that an experienced engineer would have solved in minutes. But I also solved layout problems in seconds that would have taken rounds of back-and-forth in a traditional handoff.
Third, the output is real. This isn't a prototype tool. The code Claude Code writes is production code. Paak is a real app that real people use.
What this means for the design industry
I don't think AI coding tools replace engineers. Complex systems, infrastructure, performance optimisation, security: these still need deep engineering expertise.
But for designers who've been sitting on product ideas because they "can't code"? That barrier is gone. The gap between having a vision and shipping a product has collapsed from months to days.
The designers who figure this out first will have a significant advantage. Not because they can code, but because they can ship.
Since then I've built four AI trading bots using the same approach. If you're a designer sitting on ideas because you think you need a technical co-founder, try building it yourself first. You might be surprised what a weekend can produce.
Honestly? I had a blast.
---
