We can’t AI ourselves out of a messy design system if the system doesn’t codify the context needed to clean it up.

AI will just amplify whatever structure already exists.

Well-named, well-typed, versioned design data → leverage

Sloppy, implicit, inconsistent systems → faster chaos

I think the hard part might not be building components, it’s making design intent legible to engineers, tools, and future teammates.

Tokens, schemas, APIs, docs, naming conventions; these aren’t outputs; they’re translation layers.

So, which layers are hardest to maintain? Which ones break down?

All design and engineering teams collect decisions.

The question is where those decisions live:
– in people’s heads
– in Slack threads
– in Figma comments
– or in versioned, inspectable systems

Design systems are just one outcome of codifying decisions.

Curious where this framing breaks down.