For years, I pushed back against the idea that designers should learn to code. It never made sense to me. Our time was better spent sharpening how we think, not learning syntax. But something changed recently—and it changed fast.

The ability to build what you design is obviously powerful. But coding isn’t something you pick up overnight. The risk, as I saw it, was ending up stuck somewhere in the middle: half a designer, half a developer, not great at either. Whether that’s valuable depends on the work you do. But for most designers, I still believe writing well is more powerful than coding. Being able to explain your thinking—to articulate the reasoning behind your design decisions—matters more than whether you can build the thing yourself. It always has.

But over the last few months, I’ve had to reconsider. Because now, designers don’t have to code to build. Tools like Replit, V0, Lovable, Cursor, and a dozen more are changing the rules. You can launch real apps and websites without touching code.

If you want proof, you’re looking at it. I built this site myself, using Cursor and ChatGPT. The last time I built a site entirely on my own, it was a Geocities page on SunsetStrip. (If you know, you know.)

Rebuilding my website

Anyone who’s visited my site before knows I have a habit of rebuilding it every 6–12 months. It usually starts small—an itch to tweak some part of the design—and then, before I know it, I’m redesigning the whole thing. In the past, that always meant reaching out to a developer. “Help me build it” was really just a polite way of saying “I’ll pay you to build it.”

I’d been hearing the buzz around AI tools that promised to change all that. And I tried most of them. But nothing clicked. Maybe because I wasn’t building real projects—just throwing random Figma files at them. Most tools stumbled over basic things like margins and corner radius.

I can’t even remember exactly why I decided to try Cursor instead of emailing a developer. But the moment I started, something felt different. I found myself losing track of time. I hadn’t felt that way in ages. I was waking up eager to keep working. At times it felt like a superpower: a friend who could work alongside me, ready whenever I wanted to adjust a font size or tweak the line-height for the thousandth time.

Now, don’t get me wrong. Cursor—and every other AI dev tool out there—is still far from perfect. There were plenty of moments when I (quietly) yelled at it to stop. Once it decided my site needed a second footer, complete with a different name—Anton Johansson—and links to his Twitter and GitHub. When I added my work history, it randomly claimed I had worked at Twitter. These moments are funny, but they’re also reminders of how early we still are.

And yet, even with the bugs and hallucinations, the shift is clear. The tools are getting better—fast. They’re learning to understand context. They’re interpreting design needs more accurately. Every update shaves off a little bit more friction. Not all of these tools will survive. But the future they point to feels inevitable.

What’s important is seeing them for what they are: assistants, not replacements. They’re great at the repetitive parts, at quick mockups, at handling the boring bits. But they still need you steering the ship. That balance—between AI doing the work and you shaping the work—is where design is heading.

Designers need to learn how to prompt

Which brings me to this.

I still don’t think designers need to learn to code. But I do think we need to learn how to prompt. No matter how smart the tools get, the person writing the prompt is still responsible for the outcome. Better prompts lead to better results. Every time.

And this goes way beyond Cursor. It’s true for Visual Electric. It’s true for Midjourney. It’s true for anything you’re using to create.

“AI won’t take your job. Someone using AI will.”

Here’s how I learned that.

When I first started building this site, I asked Cursor to help me choose a framework. I thought I’d use NextJS. But once I shared my real needs—static site, blog, fast—Cursor suggested Astro instead. And it was right.

Over the next few weeks, we refined the layout. Pulled blog posts in from Ghost. Connected the newsletter form through Kit. Set up an RSS feed. Even experimented with animations. (Most of those experiments nearly ended with me tossing my laptop out the window, but that’s another story.)

What I discovered is simple: the more thoughtful and specific I was, the better the work got. It wasn’t about the tool improving. It was about me getting better at asking.

That’s why I’m excited about what’s coming. This isn’t the end of design. It’s the beginning. Designers are finally getting the tools to build what we imagine—exactly how we imagine it.

So no, designers don’t need to code. But we do need to experiment. We do need to play. And we absolutely need to learn how to speak clearly and fluently with these new tools.

It starts with prompting.