Design systems are breaking. The author, a designer working inside component libraries and documentation frameworks, identifies a structural failure: guidelines written for human intuition, things like 'use this variant for emphasis' or 'avoid this in complex flows', cannot be parsed by AI tools now being handed design work. IBM Carbon gets cited by name. Don Norman's 'The Design of Everyday Things' gets stress-tested against a world its author never anticipated. The core problem is precise: machines execute, they do not interpret. They cannot infer what qualifies as 'emphasis' or 'flow complexity' without explicit logic mapping inputs to outputs.
The proposed fix is not a philosophy, it is a schema change. Component documentation needs defined input-output conditions, explicit if-then logic, and structured data formats like design tokens and component properties. The article includes a working code example, a Banner component spec with a variant resolution table keyed to 'message_type' and 'user_impact' values. That table is the argument in concrete form. It shows exactly what machine-readable documentation looks like versus what currently exists in most Figma comments and team wikis.
The piece does not resolve the tension cleanly, and that is why it is worth reading in full. The author holds both constraints simultaneously: over-specifying kills design flexibility for humans, under-specifying makes documentation useless for machines. The question of how to write documentation that serves both without collapsing into either extreme is left open, deliberately. If you work on a design system, maintain component guidelines, or are watching AI tools absorb more of the production design process, this article is mapping terrain you will need to navigate.
[READ ORIGINAL →]