Evil Martians built a design system for Currents, a test observability platform, in seven weeks using AI-assisted workflows. The project started with an LLM-accelerated audit of 791 files revealing 236 unique colors, two competing font systems, and inconsistent icon usage. The resulting design system covers typography (Innovator Grotesk with six sizes and five token groups), icons (Figura One with a 90% AI-mapped migration table), and colors (1,413 usages collapsed into four OKLCH-based token groups). A key design goal was making the system readable by both engineers and AI coding agents — OKLCH color definitions allow agents to extend the palette predictably, and token naming eliminates ambiguity. The system lives in Figma, Storybook, and GitHub. The post also outlines three signals that indicate a team needs a design system, particularly relevant for lean engineering teams using LLMs for code generation.
Table of contents
Why Currents hired Evil MartiansPart 0: designing the visionPart 1: running an AI-assisted audit in week onePart 2: building a design system specced for engineers and AITypography: Innovator GroteskIcons: a custom set, mapped to the old one with 90% AI accuracyColor: 1,413 usages collapsed into four token groupsExploring the OKLCH ecosystem and its toolsPart 3: going the extra mileResults and next stepsHow to know if you need to invest in a design system in 202647.3K Impressions1 Comment