---
id: evals
kind: topic
title: "Evals for design systems"
atlas_url: https://designsystems.fly.dev/nodes/evals.md
---

# Evals for design systems

Treating the agent-facing surface as versioned software with a regression suite: scoring, benchmarks and revision comparisons that can falsify whether it works. Measurement across a suite, not a per-change check, so it reports rather than blocks. Rare, and the clearest line between the leaders and the pack.

## Edges

- <- about <- AI UI/UX: how do you know the workflow actually worked?, Design systems need evals, State of AI in Design Systems — Insights
- <- focus <- David Cai, Kaelig Deloumeau-Prigent, Michael Shilman, Murphy Trueman, Phong Trần
- <- ships <- Astryx, Atlassian Design System, Trinity (BILL), shadcn/ui
- <- within <- Most systems ship none, and the surface is still treated as documentation rather than software., Yes. Instructions tell an agent the rules; only running it says whether it followed them.

## Sources

- https://state-of-ai-in-design-systems.netlify.app/insights
- https://blog.murphytrueman.com/design-systems-need-evals/