Rebuilding Thrive’s design system

The challenge

The design system needed rebuilding. Three libraries were in active use across the product: a legacy custom library from the original product, a partial MUI implementation, and a shadcn uplift that was underway but not yet closed out. Governance, documentation, and a clear platform standard hadn’t caught up with the scale of the codebase, which meant components from different libraries could appear in the same flows. The product was carrying the cost: navigation patterns shifted between pages, search and mobile views had drifted from the rest of the experience. The foundations needed setting properly, with the operating model to hold them in place.

The approach

Committed to shadcn with Tailwind as the platform standard, building on the uplift already underway, and put a proper initiative behind closing it out. Tight scope: component-for-component replacement, no redesign work, base tokens customised during the migration so the platform evolved toward the 2026 visual language at the same time. Partnered with Linden as principal collaborator on the token architecture and migration approach. Set up a cross-functional guild (three engineers, three designers, a lead QA, backed by the CPTO) to own the system as a maintained product rather than a finished project. Made AI baseline across the whole lifecycle.

• Held the line on tight scope so the migration closed out, where previous attempts had widened beyond their original brief
• Set up a cross-functional guild that owns the system as a product, with dedicated time protected from feature work
• Built the pattern taxonomy across four tiers (Journey, Page, Component, Element) tracking design, code, and Figma status against each pattern
• Wired Figma’s MCP server and Code Connect to flag drift between design and codebase as it happens
• Embedded AI through component creation, documentation, drift tracking, and ambient support, with designers shipping directly to production

The impact

The system is now the foundation the 2026 vision can land on. Maintained by a guild, extended by a team using AI as a matter of course, and shaped to evolve with the product rather than chase it. The hard part was never picking a library. It was building the system around the system, and that’s the part that now lets everything else move.