0:45
In this presentation, the speaker argues that design tokens have evolved from a mere foundational tool to the core vocabulary that AI agents use to understand and manipulate UI components. He emphasizes that every visual property—colors, spacing, radii—must be expressed through tokenized values, and that these tokens should be organized in a clear hierarchy, ranging from low‑level primitives up to component‑specific tokens. Crucially, token names must be semantic, encoding the intent and context of their use (e.g., avoiding opaque names like "blue‑100"), so that AI systems can correctly interpret them. When semantic naming alone is insufficient, additional routing mechanisms are required to disambiguate token usage for agents.
