Developer builds machine-readable design library to close AI's visual style gap

A developer has created Curio, a library that translates real-world design styles — such as Swiss International Style, Bauhaus, and Memphis — into structured, machine-readable files that AI models can directly consume. The core problem it addresses is that large language models understand design vocabulary but lack grounded, consistent representations of what specific visual systems actually specify in terms of grids, type scales, spacing, and rules. Each style in Curio is packaged as a DESIGN.md file containing tokens, component specs, and explicit do-and-don't rules, giving AI a fixed set of values to work within rather than forcing it to guess. The catalog is built primarily for AI agents rather than human browsers, featuring an llms.txt, a full style index, and an open JSON catalog so agents can discover and apply styles autonomously. The developer argues this agent-first publishing pattern — pairing llms.txt with open structured data — is broadly useful for anyone building products that LLMs are meant to consume.