Lab · three pieces
Design systems for humans and agents
Experiments in React, AI-assisted prototyping, component systems, and interfaces where design decisions survive all the way into code. Self-directed, on synthetic data, and built end to end: each one is a question about how a product team works once coding agents are on it, answered with something that runs.
-
Agent Review: making a design system executable for coding agents
If agents can write production UI, how does a team make sure they inherit the decisions already made? A React component system with tokens and a Storybook, the rules an agent reads before touching it, a controlled experiment where Claude Code added a feature inside it, the validation that caught where it drifted, and the tool a person uses to review the consequences and send a precise correction back.
-
A broken deploy the developer can read and roll back
A multi-service deploy console for the moment something broke, arranged in the order developers ask their five questions: what is live, what failed, what changed, what depends on it, can I put it back. The deploy serving traffic kept apart from the newest attempt, a failed build explained on one card, and a rollback previewed before it runs and undone in one press.
-
An AI recommendation the dispatcher can check and overrule
Human-in-the-loop decision support for freight dispatch: the factors behind a pick shown with their values and their age, a close call that names its tradeoff rather than picking, a legal limit the system enforces before it scores, and a one-click override that asks why afterwards. Built on one hypothesis: a dispatcher spends less time reconstructing a recommendation when the interface shows what needs checking.