Things I've written
I write about design engineering, UX research, design systems, accessible React, and AI-integrated product design.
Developer tools
A library is a product with an install step, a README, and an API surface. These essays treat that experience as interface design.
Featured
An agent’s check passed. The review is still yours.
Building a bounded permission and review flow with Primer React, an explicit state model, and inspectable evidence.
How I use Strands Agents to make Waco3 tutorials
A reviewed script becomes a screen recording, local narration, and an edited walkthrough. The workflow, review boundaries, and TypeScript scaffold.
The Best AI Workflow Still Ends With a Human Review
AI gets a lot right. It also misses things a human catches in 30 seconds — edge cases from domain knowledge, whether the solution actually solves the problem.
AI Interfaces Need Receipts
When AI makes a decision, the interface needs to show the reasoning — not as a disclosure, but as primary UX. Trust comes from transparency, not magic.
Five Themes, One Pipeline
Brutalist mode and high-contrast mode have opposite goals and ship from the same token source. The layered architecture that makes multi-theme a naming problem, not a CSS-forking problem.
What a Combobox Teaches You About Product Quality
Building a WCAG-compliant combobox forces keyboard, screen reader, and edge case decisions that reveal how much you really know about your UI.
Agents Should Show Their Work, Not Their Chain of Thought
Chain-of-thought traces are often wrong. Real transparency means showing the files read, commands run, and decisions made — not the reasoning monologue.
The Screenshot Is the Missing Test for AI-Generated UI
AI code can pass unit tests while producing visually broken UI. Visual regression testing closes the gap — here's how to wire it into an AI coding workflow.
Context Engineering Is Just Information Architecture With Consequences
What you put in an LLM's context window is information architecture. The same principles apply — priority, grouping, hierarchy — but the stakes are higher.
A Skill Is Better Than a Prompt When the Work Repeats
When to encode a task as a Claude Code skill versus writing a prompt each time — and how to write a skill that actually holds up.
AI Coding Agents Need Threat Models Too
Prompt injection, credential exposure, supply chain risk, scope creep — security considerations for AI coding agents and the specific mitigations that work.
The Agent Permission Model I Want in Every Coding Tool
What Claude Code's permission system gets right, what granular agent permissions should look like, and why ask-everything vs full-autonomy is a false choice.
MCP Is Boring in the Best Possible Way
The Model Context Protocol is infrastructure that works by not being interesting. It solves real problems without drama, and that's exactly why it matters.
Motion Graphics Made Me Suspicious of Pretty UI
What a background in After Effects and broadcast production teaches about product animation — and why most UI motion is noise.
The Interface Should Tell You What Just Happened
Feedback patterns in UI — how interfaces communicate the result of user actions, when toasts fail, and what to use instead.
Scroll Animation Is Usually a Content Problem
Scroll-triggered animation often hides weak content. Before adding another reveal animation, ask whether the content actually earns it.
Microinteractions Should Have a Job Description
Microinteractions that communicate state earn their place. Ones that exist purely for feel are noise. Ask: what does this communicate, to whom, and when?
The Timeline Editor Taught Me How to Design State
Timeline editors like After Effects and Lottie have explicit state machines. Designing one taught me more about UI state than any React tutorial.