Design Systems / AI / Generative UI
A design system an agent can build from. A search box that turns intent into the right interface.
Enaia · Design systems · Design & engineering
A system for people and agents
Enaia’s design system is built to be consumed by an AI agent as much as by a person. One authored contract per component generates documentation, a component registry, token bindings, and a Figma plan. A malformed contract fails the build.
Rules carry IDs, tiers, and detection methods, so a review produces a verdict that can be traced back to the system. The generative UI demo brings that work into the product: a search box that turns intent into the right interface.
Measuring what the system changes
I built the system, then built the instrument to find out whether it actually works. Three screen-building tasks, given to Claude Sonnet 5 five times each, in three environments: with the design system, with its components but no written guidance, and with nothing.
45 runs, $41. Every result comes from an append-only, hash-chained journal, so the numbers can be recomputed.
- 49% fewer design decisions per element with the system: 1.44 against 2.82, ranging from 13–77% across tasks.
- Half the styling vocabulary: 19 distinct CSS classes per screen against 40.
- Five runs of the same prompt used 27 distinct styling decisions with the system, 114 without.
- Zero governed components reimplemented by hand across fifteen runs. Without the system, thirty-three.
The result that went against me: my written guidance initially measured as doing nothing. A grader that reads composition found where its effect had been all along.