I've always been a builder.

Long before "director" showed up on a business card, I was a kid on my family's farm in Connecticut, planting Christmas trees under the hot sun. Years later I built an app that people still use to find their way around those same fields. I wrote it myself in Swift, gave it away, and it managed to hold a five-star rating for ten seasons running. That instinct has never left me. In every job I've had, the part I quietly guarded was the part where I'm actually making the thing.

Maple Row Farm. Where the building instinct started.

 
I've led design teams for a few years now and I'm good at it, but the work I'm proudest of was never the org chart. It was the thoughts contained in the file. An agentic interaction model. The three taps I cut between a person and the answer they came for. The features I persuaded the team to not ship. I think best while getting my hands dirty, and lately that's precisely where I want them.

Which brings me to what I actually want to talk about: what's happening to our craft right now.

The floor under the whole UX industry has just dropped significantly, and it's without question the most interesting thing to happen to design since Gutenberg. Generative AI made baseline production almost free. Developers get through tasks something like 55 percent faster with it, roughly a quarter of production code is now machine-written, and close to nine in ten designers reach for AI in any given week. Anyone can spin up a clean prototype, or a serviceable paragraph. Producing artifacts used to be the moat. It's becoming the price of entry.

When making things becomes this cheap, you don't get clarity. You get a pile of AI slop. Artifacts everywhere, assumptions quietly diverging, no shared problem underneath any of it. So I stopped treating a prototype as the deliverable and started treating it as an instrument. Each one I build is really a question about how the system should behave, not a picture of how it should look. The skill now is aiming that cheap production energy at the questions worth asking instead of constantly feeding the pile. When everyone can produce artifacts, the value quietly moves to knowing how they can be stitched together, if they even should be, or which of them are even worth producing at all.

This is self-evident in the number that refused to move. For all this massive AI adoption, organization-wide productivity has barely shifted. It's hovering around ten percent, because the real bottleneck slid downstream into judgment, review, integration, taste. The floor came up, but the ceiling hasn't moved yet. And the ceiling can still be just one person with a strong, defensible opinion about what good looks like. That gap between cheap production and scarce judgment has never been wider, and it's the part of the work I like the most.

The floor rose while the ceiling hasn't. The gap between them is where the satisfying work is.

I work as an AI-native designer by default. My intelligence stack is the set of tools I build with. It organizes a complex, end-to-end AI workflow into modular, interconnected layers, from ideation to user-facing applications, and it's the consistent thing I build through on every project, from first idea to shipped screen. Claude Code and Figma MCP take me from concept to a real, working prototype with no handoff, so I'm pressure-testing interaction models in days, sometimes hours, against a live design system instead of waiting weeks for a build. I design recurring agents, the persistent workflows that handle repetitive cognitive work on a schedule rather than sitting idle until someone asks. And I lean hard on agent design review, because in an agentic product the thing you're shaping isn't a screen, it's a behavior: what the agent assumes, when it checks with you before acting, whether it fails out loud instead of quietly, how it earns a little more trust with each successful pass. Someone has to hold the bar on that behavior, and I think it has to be a designer who understands both the model and the human on the other side of it.

There's a bigger shift tucked inside all of this. As people stop doing tasks by hand and start handing them to agents, the work turns from designing a user experience into designing a manager experience: how someone delegates a task, watches it run, steps in to fix it, then recovers when it goes sideways, and decides how much to trust it next time. What I'm designing isn't a screen anymore, or even a single agent. It's the whole system of people and agents and the ways they push on each other.

It's also why design keeps absorbing more of the work. Once a designer can build, the handoff simply collapses. What used to take three roles and three weeks can take one person and an afternoon. Design not only stops specifying the product and starts producing it, but it picks up scope that used to sit with engineering and product. Nobody handed us that permission. The tools just finally made it possible.

What I care about is covering both ends of that gap at once. Speed by itself is a commodity now, and counting prototypes is just counting speed dressed up for the dance. The version that ultimately matters is using the stack to move fast and bringing hard-won judgment to keep everything high quality because the judgment is the only truly scarce part of all this. Like a chef who both works the line and runs the pass. The speed of cranking out dishes and the judgment of deciding which ones leave the kitchen, held by the same hands.

I've got the receipts. I've built agentic interaction models that shipped and upleveled a team to design the same way. I've endured contested calls that ultimately pushed search use up 200 percent. A platform NPS that climbed from 20 to 48. A small app I built alone that's outlasted every framework war of the past decade and still earns five stars. Velocity and quality, both actually shipped.

I'm most at home where the ceiling matters and the bar is high. That's the work I want to keep doing.

Thank you for reading this.