Aurora Core is the design language that ties together every personal project I’m currently working on. It’s packaged as a production-grade foundation kit rather than a static handoff — token exports in five formats, 25 platform-specific reference implementations, a Figma library plan, Zeroheight article templates, governance docs, and a set of FastMCP servers so future codegen flows can query the system directly.
Why it exists
I wanted the independent work I do at night to hang together visually, even when the pieces are native desktop apps, web apps, and CLI tools. Ad-hoc color and spacing choices add up fast across a portfolio, and I wanted an AI-friendly source of truth — not just a style sheet — so agents generating components can do the right thing without supervision.
Highlights
- Seven anti-pattern rules that govern meaning: gold glows, amber is matte, purple is analytical, blue is informational, depth is blur-driven, tracks inherit semantics, toggles live in the secondary spectrum.
- Canonical sampler is a single self-contained HTML file — every component renders live, in both dark and light themes, without a build step.
- Token exports in five formats: JSON, W3C draft, Tokens Studio, CSS custom properties, and a Tailwind preset.
- FastMCP servers expose the system as tool calls so Claude Code and Codex can introspect tokens, fetch component snippets, and validate their output against the design language.
- Reference implementations across web (6), mobile (6), desktop (10), plus Storybook and Tailwind integration.
Technical approach
The authoritative source is a single HTML file with inline CSS and a small JS bundle for interactive demos. Everything downstream — Figma variables, Tokens Studio, W3C exports — is derived from it through export scripts rather than maintained separately. This keeps one source of truth and makes it trivial to verify a token across every consumer.
