Design at AI Speed

Three weeks. One demo. A designer in the terminal.

I was listening to a recent episode of the Lenny Podcast — an interview with Jenny Wen — and a few things she said landed differently than usual. They resonated with what I'd been living through on one of my recent projects.

Let me tell you what's been going on.

The Setup

I've been working on a product in the hardware domain, and the first feature I touched is a different beast. The complexity is real, the constraints are tight, and the stakes feel higher when things get physical. On top of that, our timeline is compressed. Every week counts.

My most recent challenge: a highly complex data sheet that pulls from multiple APIs, while the developer is building the backend in parallel — at the same time I'm designing. No waiting. No handoffs. Just two tracks running simultaneously, racing toward the same demo.

So the question became: What do you do when there's too much to do and not enough time?

What I Actually Did

I stopped treating design and development as separate phases.

1. Understand What I Have

Context matters. I inherited what existed in Figma, cleaned it up, aligned it to the design system, and consolidated the foundation before branching out. Then I did a quick product discovery around three questions: Why now? Who's the audience? What's the goal? Understanding the vision — but being practical about what the demo actually needed to ship.

From there, I identified the two major challenges: an unclear user flow tangled up with ongoing backend work, and a confusing navigation structure running three to four tiers deep. Both affected each other. Both needed to be solved.

2. Parallel Exploration

Rather than resolving the ambiguity before starting, I moved on both fronts at once.

In Figma, I built component structures that matched the data model — flexible enough to evolve as the API shape changed. I started with a quick information architecture to map statuses and flows, then iterated with stakeholders through validation sessions to pressure-test the direction.

At the same time, I used Claude Code to prototype the navigation directly in code. Real interactions. Real data shapes. No faking it. Instead of waiting to "validate" ideas before building, I was exploring and discovering in the medium that mattered — which meant the frontend conversation was grounded in something tangible from the very beginning.

This is where AI changed everything. The speed of going from idea → working prototype collapsed from days to hours.

3. Convergence, Not Handoff

By week three, we had two things running in parallel: a high-fidelity Figma system and a code prototype stress-tested against real complexity. Bringing them together wasn't a handoff — it was a convergence. We folded both into a production-ready design the developer could build directly from. No translation layer. No loss in meaning.

T-Shaped in an AI World

This connects back to what Jenny Wen was talking about. The designers who thrive right now aren't the ones who are precious about their lane. They're T-shaped, deep in craft as an IC, but wide enough to move fluidly into adjacent territory: prototyping, front-end, systems thinking, data.

AI makes the width of the T easier to develop. The barrier to picking up new tools and techniques has dropped dramatically. What used to take weeks to learn enough to be useful now takes days.

A few things I'd add from my own experience:

• Be a quick learner, not just a fast worker. Speed without learning is just rushing. The goal is to absorb new domains fast and apply them immediately — not to cut corners.
• Don't let tools define your limits. Whether you're working in Figma, an IDE, or a terminal — and increasingly, designers working with AI are in all three — the designer who understands what's theoretically possible and then uses the technology to get there is the one who ships things others said couldn't be done.
• The bottleneck has shifted. With AI as a collaborator, execution is no longer the constraint. Imagination and judgment are. Jenny Wen put it well: the design process as we knew it is dead. What remains — and what actually differentiates — are the people who know the product deeply enough to make the right calls, and the people who are humble enough to keep learning fast. That combination is what the best teams are hiring for right now.

What I'm Taking Away

We're in a moment where the definition of "designer" is expanding faster than most job descriptions can keep up with. The work I did in three weeks — straddling Figma and code, design and data, exploration and production — would have taken a much larger team not long ago.

Somewhere along the way, I stopped thinking of myself as purely a designer. I was in the terminal. I was reading API responses. I was making decisions that touched both the system and the surface. AI made that possible, but the curiosity had to come first.

That's not a threat to what design is. It's an expansion of what a designer can be.