Why I upgraded three sites in the same week
I had three client projects on Tailwind 3.x that all needed new features in the same sprint. Tailwind v4 was out, the new engine looked fast enough to justify the risk, and I did not want to maintain two different utility dialects for long.
So I blocked a week and upgraded all three in one go. This is what actually mattered: the utilities that kept working, the custom pieces that did not, and the order that made it survivable.
The order: simple static, then shared design system, then the weird one
The main decision was to pick an order that let me learn on a safer codebase and reuse that knowledge on the more fragile ones.
1. The simple marketing site
First was a small marketing site: no complex build pipeline, no design tokens, just Tailwind plus a few components.
The goal here was not perfection. The goal was to get a green build, scan the UI, and see what Tailwind v4 does to a real project without a lot of custom layers in the way.
2.…