One philosophy. Four phases.

Philosophy

• — "Less is more" — every extra token costs ~30 variables per theme.
• — "Foundation is not a technical simplification — it's a product decision."
• — "Overrides are allowed — but they're the last resort."
• — "Dark mode by construction, not by override."
• — "The system must be understandable by designers, not just engineers."

Four founding decisions: semantic naming (not by value), layer separation, color decomposition, and accessibility algorithm. The Alpha had no dynamic generation — everything was manual. But the conceptual decisions from this phase are the DNA of the current system.

V1 formalized depth, opacity, typography, and borders. It introduced the concept of Engineering Summaries — technical documentation of the rationale behind each system decision. V1 also identified the limits: nomenclature still carried legacy from Alpha, and token generation was partially manual.

V2 resolved V1's limits with eight innovations: dynamic generator, OKLCh pipeline, Dimension layer, canonical naming, Foundation as composed styles, gradients, migration bundles, and multi-platform outputs. The system became deterministic — given the same config, the output is always the same.

V3 made the system consumable by any project. @aplica/aplica-theme-engine is published as a standalone npm package — configuration and data live in the consumer project, not the package. The architecture supports any component library and any token pipeline, connected via AIs, parsers, and tools. The AI Skills Injection Program (ai:init) distributes contextual guidance directly to Cursor, Claude Code, and GitHub Copilot.