While the industry debates whether AI will replace designers (most likely not) and developers embrace “vibe coding,” I was learning a different lesson: the problem isn’t whether AI can do the work—it’s figuring out which work actually needs doing.
After creating all those foundation documents for my redesign, wireframes seemed like the obvious next step. You know, follow the process, do things properly.
Except I was about to learn that “properly” isn’t always right when it comes to inserting AI into a more traditional processes.
The Control Problem
My logic seemed solid: without wireframes, each page would look different. Fine for an art project or a Geocities throwback (kids, ask your parents), but this was my professional presence. I needed control and coherence.
Maybe it was my prompting skills, or lack thereof, but every tool I tried felt like fighting uphill. Despite what the marketing promised about “just describe what you want,” reality was messier.
The Tool Graveyard
Claude’s SVG Experiment
Did you know Claude can make SVG files? The process was slow and heavy. It would say it fixed something, but didn’t. Artifacts work great for documents, but wireframes need to be seen side by side—impossible when each layout is trapped in its own little box. I also think that, creating consistency between the artifacts is like conducting an orchestra where each musician is a cat.
The Canva Mystery
I tried Canva next, but honestly never understood the interface. It’s like my old Photoshop trauma all over again—you know, when you just want to crop something simple and three hours later you’re somehow in the Channels panel wondering how you got there, ultimately giving up and opening MS Paint.
Maybe Canva is perfect for wireframing and I’m just dense.
Relume: So Close
Relume was the closest I got. Give it prompts and it builds a whole sitemap, wireframes, and style guide. Pretty impressive!
The catch? Relume couldn’t use those foundation documents from part 1 as-is. I had to summarize everything to fit into their prompt interface. A better feature would’ve been importing markdown files directly, but that’s wishful thinking.
But I’m also cheap and couldn’t justify $18 USD ($25 CAD) for another tool. The templates all felt Nike.com-inspired. If I’d continued, I would’ve exported the React components into Cursor, but that felt disconnected from everything else.
I sat on it for weeks, tweaking and never happy.
Hey! How about you ask someone what they did?
Chatting with a friend about his own AI projects, I asked: “Did you use wireframes? What was your workflow?”
His answer: “For simple projects, I just jump straight into coding.”
Lightbulb moment. While everyone’s trying to AI everything, maybe I was overcomplicating this.
The “Why Didn’t I Think of This Sooner” Moment
I was solving for problems that didn’t exist:
Consistency: I already had a design system. That was the consistency framework.
Professional process: This wasn’t client work. I could be more experimental. Maybe not contemporary dance experimental, but definitely more experimental than my usual “measure twice, cut once / analysis paralysis” approach.
Control: Sometimes the best way to understand how something should work is to build it.
Those foundation documents I’d created weren’t just planning—they were my wireframes. They provided the framework I needed without translating every page into detailed layouts first.
What’s the Lesson Here, Kiddo?
AI tools don’t make every traditional process step necessary.
For simple projects, especially solo work with clear design principles, you can skip wireframes and prototype directly in code. The foundation documents provided the framework—I didn’t need to translate every page into detailed layouts first.
Sometimes questioning the “obvious” next step is the most productive thing you can do. Maybe the future isn’t about better AI tools, but better AI collaboration.