Justin Lokitz invited me onto Designshift to talk about something I'm still working out in real time: what happens when a twenty-year design leader puts down the team and starts building things alone.
The full conversation is below. Three things kept surfacing for me in the days after we recorded that I want to sit with here.
Blue-collar problems
The thing I keep returning to is this: I'm not trying to define a category. I'm working on what I call blue-collar problems — the unglamorous, real-world stuff that's been sitting in someone's life for years. A cyclist who can't remember when she last replaced her chain. A youth-sports coach managing twelve parents over text. A daughter trying to coordinate care for an aging parent across three siblings.
These problems don't make pitch decks. They make products people actually use.
It's a deliberate counter-positioning. Most of what I see in the AI-product space is chasing the next big surface — the new agent paradigm, the new interface metaphor. I'm interested in the boring middle: take a problem people already have, and use the new tools to solve it well.
The design work didn't change. The execution did.
The thing AI hype gets wrong is that the design part of design is mostly unchanged. Framing the problem, mapping the real constraints, deciding what matters and what doesn't — those still take experience and judgment. What collapsed is the distance between a decision and a working artifact.
For most of my career, the gap between "we should build it this way" and "the thing exists" was measured in sprints, headcount, and political capital. Now it's measured in a focused afternoon. That's a real change. But the upstream work — the part where you decide what to build and why — looks almost identical to what it did ten years ago.
If anything, the new tools have made me more careful about that upstream work, not less. When you can ship the wrong thing in a day, you become very interested in being right.
Judgment is the new bottleneck
Most problems aren't actually that complicated once they're framed correctly.
I said this to Justin and it landed harder than I expected — because the rest of the conversation kept circling back to it. As the cost of building approaches zero, the cost of building the wrong thing is the only cost left. So the bottleneck moved upstream. It's not "can I build it." It's "should I, and what is it actually for, and who is the one person whose life this needs to change."
That work hasn't gotten any easier. I wrote about this here too — the short version is that the design leaders worth hiring right now are the ones who've been in the room when things went sideways. AI will get faster. Judgment is the part that takes twenty years.