Every week, it feels like there's a new AI tool promising to make my work "faster" or "easier." And every week, I find myself a little bit skeptical and a little bit curious.
That tension came to a head last week when I read an article arguing that designing in Figma might be outdated, replaced entirely by vibe designing directly in code through AI. It's a provocative take, and honestly, a very interesting one. But I’m not convinced yet.
“Figma Never”
The article I read leans into a compelling question: why mock something up in Figma when AI can just build it? But I think this reduces Figma to pixel-pushing and pretty pictures of software. That's a narrow view of what design tools actually do.
Figma isn't just an aesthetics tool. It's a canvas for exploration, iteration, and collaboration. It holds space for uncertainty in the design process: for the messy, abstract early stages when the problem isn't fully formed and the story is still taking shape. That kind of open-ended thinking is hard to replicate when you're committing to code.
Fully functioning prototypes in code is powerful, but it's a storytelling tool best used when the vision is mostly solidified. When you want to immerse stakeholders in something that feels real, interactive, and close to the final product, code-first design shines. But when you're still figuring out what you're even building? Jumping into code too early can close off exploration before it's had a chance to fully bloom.
There's also a practical issue: AI tools depend on user precision when early design thinking is anything but. It's hard to perfectly describe a color palette or an interaction style in a prompt, and when AI misinterprets you (which it will), you end up feeding it references, tweaking outputs, and rebuilding things in Figma anyway.
Skipping Figma doesn't always save time. The most efficient process will likely include both, using the right tool to tell the right story at the right time.
Exploring AI Tools: Subframe
With that philosophy in mind, I pushed myself to actually test one of these new AI-powered design platforms. With encouragement from @Olivia, I spent about nine hours exploring Subframe, a design software tool positioned to bridge the design-to-engineering gap (using its Pro Plan at $29/month).
My project: a concept for a coffee shop POS system, inspired by an adorable café I'd visited that morning. A low-stakes, real-world use case that I could relate to as someone who spent a good amount of time in the food service industry.
The text-to-UI generator surprised me immediately. From a single prompt, I had four layout options in under a minute. The visuals and typography needed adjustments, but the layout logic was solid for barista workflows. That kind of structural thinking, generated that fast, is genuinely impressive.
Editing was where the friction crept in. Subframe's design tools felt limited compared to Figma. Customizing colors, typography, and spacing was more constrained than I'm used to, and adding complexity to prototypes got frustrating quickly. That said, with more time I’m sure designing in Subframe would eventually become more comfortable.
But here's the thing: that simplicity might actually be a feature for non-designers or developers who find Figma overwhelming. Not every user needs granular control.
The biggest feature to me was the component library and its real-time React code updates. Watching design decisions immediately reflect in production-ready code felt like a glimpse of a better design-to-engineering handoff. That's the kind of friction that costs teams real time, and Subframe tackles it directly.
My Takeaways
I'm not ready to close Figma. But I'm also not dismissing tools like Subframe. What exploring them has reinforced for me is that we're not in an either/or moment right now, but rather a when and why moment.
AI tools are getting good at the repetitive, structural parts of design. That should free designers to focus more on the creative and strategic parts of the process that are still messy and human (and fun!) and hard to prompt your way through.
The more designers engage with these tools, the more influence we'll have in shaping how they evolve to actually serve the design process rather than just accelerate it.