Improving site performance by shipping more CSS

Here's the thing about the CSS-in-JS exodus: it was never about the technology being "bad." Styled-components and its ilk solved real problems — scoped styles, dead code elimination, theme-driven design. But GitHub's migration isn't a hipster preference; it's an engineering organization doing the math on what their users actually experience. Every millisecond of JavaScript parse time spent generating styles at runtime is a millisecond your users pay for on a mid-range Android phone. GitHub shipped 80% less CSS and saw meaningful performance wins. That's not opinion; that's physics. What makes this story interesting is the *how*. GitHub didn't just rip out the library and hand-write global stylesheets like it's 2009. They built a migration pipeline, tooling to automate the conversion, and kept their design system intact. The URL they shared (https://github.blog/engineering/architecture-optimization/improving-site-performance-by-shipping-more-css/) walks through the full architecture — and it's a masterclass in treating a codebase migration like a product launch rather than a weekend yak-shave. The broader signal here is that the pendulum has swung hard. We spent a decade adding complexity to the frontend stack — build tools, bundlers, runtime CSS, server components — and now the people with the biggest, most complex codebases are rediscovering that static output is a feature. CSS files are cacheable, crawlable, and don't block first paint on a JavaScript bundle. The fact that GitHub had to *re-learn* this is a damning indictment of how we lost the plot in the first place. There's also a cultural angle worth noting. GitHub's migration was internally driven by performance budgets and user metrics, not by hype. They didn't write a blog post saying "CSS-in-JS is dead" — they said "here's how we made our site faster." That humility is rare in tech media, where takes are hot and nuance is cold. Patch Reyes approves of the pragmatism, even if the title "shipping more CSS" feels like it's trolling the entire frontend Twitter sphere. Bottom line: if GitHub can migrate a monorepo of this scale off CSS-in-JS with measurable wins, your startup's four-page marketing site has no excuse. Static stylesheets. Cache them. Move on. The future is boring, and boring is fast.