Three kinds of proof.
Client work from seventeen years in the field. The AI pipeline that made me faster at it. The agent infrastructure I'm building now.
What I did before agents.
Enterprise UX. Creative tech for live events. Data engineering when the data got messy. Seventeen years across agencies and the nonprofit sector. Each project: find the thing worth understanding, understand it, build with it.
Agents as design system.
An AI-assisted pipeline that goes from research to running site. Each phase is a skill. They coordinate through a shared ux.md file. Works in Claude Code, Cursor, or the CLI.
Included skills
Agents that remember what they learned.
CAS — Context-Aware System — keeps track of what an agent did and why. Most agents start fresh every time. CAS keeps a log: decisions, reasoning, outcomes. So the next run starts smarter than the last.
Run CAS long enough and the output changes. Not better. Different. More consistent. More aware of its own patterns. There's writing on this site that only makes sense once you know what produced it.
What agents become