“The thing that kills a design system is not missing components. It is the gap between the system and the actual product decisions being made.”

“I’ve watched design systems rot because they were treated as Figma documentation instead of infrastructure. At Eccho, I started treating tokens and components as executable contracts that compile to code, because once LLMs touch the system, soft guidelines collapse.”

“The thing that quietly kills a design system is when product teams start treating it like a library they check out from instead of a shared thing they have to keep alive. I spent a lot of time at Mark43 and Ramsey Solutions watching components drift into a dozen local variants because the system wasn't answering a real question the team had that week.”

“The systems that survive are the ones people actually want to use. With Arabesq, I saw that 1.3k+ users stayed because the components solved Arabic layout pain, not because the docs were perfect. I now spend more time making the system the path of least resistance in Figma and code than writing rules no one reads.”

“I’ve seen too many teams mistake a Sketch UI kit for a design system and then wonder why it crumbles the moment another product needs something different. The thing that holds it together is treating it like a real product: a dedicated team, a roadmap, a budget, and a loop where pilots feed back into the system instead of drifting off on their own.”

“The thing that kept Meridian alive after the team was disbanded was that we made it easier to use the system than to avoid it: 850K+ Figma instances, direct links from components to code docs, and variables like size-sm so designers and developers speak the same language.”

“The thing that kept my design systems alive was treating the system itself as a product with real users: the other designers and engineers. I spent too long building components nobody asked for, then watched them build their own anyway.”

“At CANDIS, I built our design system as a single person design team because engineering was scaling faster than design could keep up; the system stayed alive only because I treated it as a product with its own users, the developers, and shipped changes they could adopt without asking me.”

“I merged three visual languages into one component library at Happly, and the part that actually mattered was the boring part: running design QA across every surface in both languages and both tenant colors.”