About — the pivot
Designer first. Engineer when it helps.
Ten years of product design. Now writing the code too, with AI as the pair. The how and the why.
Then
I’ve always been a designer who codes. The medium has always been the point — same way a graphic designer understands paper and printing. Right now we’re moving through another transformation, and it asks for an understanding of a new medium to design in a new way.
What carries forward: good design still applies, especially now. AI is a force multiplier; it amplifies the precise and the imprecise equally. The specifications need to be tighter, the understanding deeper, the guardrails more careful. Design skills don’t go away. They get documented, leveraged, enhanced.
Catalyst
I’d been watching the shift for a few years. Early 2026 was the moment it became obvious — time to change the foundation of how I work and design.
Coming back to my roots: a designer who codes, this time paired with AI. The iteration cycles are different. Ideas I couldn’t have afforded to test a year ago can land in an afternoon now.
Now
As a product designer for AI-native workflows, every layer of the practice is in play at once.
Designing for trust— when the model isn’t sure, the interface says so and the person decides what happens next.
Streaming as design material— the AI’s output is something to shape, not just print in a chat bubble.
Knowing when AI hands back— when the system can’t take the next step on its own, the design makes that handoff obvious.
Why this combination
Ten years of taking products from zero to one — finding the right thing to build, cutting what doesn’t ship, getting a working version into users’ hands. The work trains a specific kind of attention: the user is not the system, and the system is not the user. The user has their own job. The tool serves the job. The design work is keeping that gap visible.
AI-paired coding closed the last loop. I direct the model, read what it writes, redirect when it’s wrong, and iterate on the real interface — not a static mockup someone else has to translate. The thing I design is the thing I ship. No handoff. No game of telephone with an engineer who half-gets the intent.
Anyone can prompt an AI to write code. The rare part is knowing what the first version should be — what to cut, what to keep, what’s worth shipping now. That call doesn’t come from a model. It comes from ten years of doing it.
How I work
Operating principles
- Ship in code
- Every design is the actual interface. No mockups handed off to engineers.
- Real users first
- v1 in front of users before the v2 conversation. Cuts the speculation early.
- Documented intent
- The spec is the ceiling on the output. Concrete details — padding values, color tokens, interaction patterns — push the ceiling up. Anything left to interpretation comes back as drift.
- Operator framing
- The user has their own job. The tool serves that job. Design keeps the line between the two visible.