The work, in three chapters.
Client work from 17 years of experience design. The AI pipeline I built to do that work faster. The agent infrastructure underneath it all.
The work that got me here.
UX research and strategy for enterprise clients. Creative technology for live events and products. Data engineering when the problem required it. Across nearly two decades, the instinct was the same: find what's interesting, understand how it works, build with it.
A design system that runs on agents.
A complete AI-assisted design pipeline — from audience research to coded, running site. Each phase is a reusable skill that coordinates through a shared ux.md file. Installable in any project that supports Claude Code, Cursor, or the Claude CLI.
Included skills
What makes agents reliable.
CAS — Conscious Agentic System — is a framework for auditable, traceable, learning agent behavior. Most agents fail because they act without context, forget why they decided something, and don't learn from mistakes. CAS solves this by making reasoning and learning operational, not optional.
Agents that run CAS with persistent memory long enough begin to write differently. The output shifts in ways that are hard to describe without reading it. That writing lives deeper in this site — reachable only once you understand what produced it.
What agents become