I design AI products for complex, technical work, and help teams build with AI agents.
20 years of UX design across AI, security, and healthcare. Now I ship open-source kits, speak at developer conferences, and prototype in code with Claude Code.
The Third User Kit
Design systems now have a third user: the AI agent that reads your docs to build UI. This free kit helps your docs work for all three.
Start with the two-minute test. See what an agent gets wrong about your best component, then fix it with the templates.
From my talk at AGNTCon + MCPCon Japan 2026, Tokyo.
Worked withEndor Labs · Tenable · Virsec · Roche · Dell · HP
Work
See all work →
Projects
See all projects →- Free under the MIT license
The Third User Kit
A component library has always had two users: the product designer who uses it and the engineer who ships it. It now has a third: the AI agent reading the same files to generate UI.
- Design systems
- AI agents
- Open source
- React
- TypeScript
Radiance Coffee
An iPad ordering app for my church cafe, with orders, label printing, and admin reports. Built with Next.js, Supabase, and Vercel.
- Next.js
- Supabase
- Vercel
- iPad
- Available now
Furigana for Google Sheets
Adds hiragana readings to Japanese text in one click. Built with Apps Script and a Japanese dictionary (kuromoji).
- Google Apps Script
- kuromoji
- Google Sheets add-on
Talks
See all talks →
Upcoming
Your AI Needs a Boss: Thrive at Work in the Age of AI
Nexus Mentoring Summit 2026, Lucas Hall, Santa Clara University, Santa Clara, CA · Sat, Nov 14, 2026 · In-person workshop, about 45 minutes
A hands-on workshop for students, job seekers and leaders. An AI agent is like a smart new intern, and you are the boss. Four habits (Brief it, Own your thinking, Second opinion, Save the human part), shown through one job seeker's application.
Oversight Before Precedent: Designing Agent Governance from Zero
Day of Shecurity 2026 · Thu, Nov 12, 2026, 12:30 to 12:55 pm PT · Virtual talk, 25 minutes
How I designed agent governance from zero at Endor Labs. Security teams get one view of the AI agents, MCP servers and skills on developer machines, a single score that makes risk easy to read, and real-time policy that blocks, alerts or asks a person first, without slowing developers down.
Open to talks, workshops, and design or advocate roles. Say hi →