Every design system I've worked on has gone through the same phase: a burst of energy where the team adds components fast, because momentum feels like progress. And every one of them eventually hit the same wall, a library with four different "close" icons, three card variants nobody could tell apart, and a Slack channel full of "is there already a component for this?" questions that took longer to answer than building the thing from scratch would have.
The instinct to treat a design system like a product catalog — more items, more coverage, more done is understandable, but it optimizes for the wrong thing. A system's job isn't to contain every possible interface. It's to be a shared language specific enough that two designers who've never spoken can build something that feels like it came from the same team.
What actually built trust
Across the product lines I've maintained component libraries for, the useful discipline was almost never "add more." It was:
Saying no to near-duplicates. Before adding a new component, the better question is whether an existing one can flex to cover the case, a new modifier is usually cheaper than a new component, and it keeps the system's mental model small.
Deprecating on purpose. Unused variants don't just sit there quietly, they actively cost trust, because every new person has to figure out which of the five button styles is actually the current one.
Naming for the reader, not the author. A token or component name should make sense to someone who wasn't in the room when it was created. If it needs a Slack thread to explain, it's not documentation-ready yet.
A system that grows without governance ends up as confusing as no system at all, just with better branding.
The real test
The best compliment a design system gets isn't "it's comprehensive." It's when a new designer or engineer joins a team, needs to ship something in their first week, and doesn't have to ask five people what a token means to do it correctly. At that point, the system has stopped being a project people maintain and started being infrastructure people simply build on, which is the whole point.
