Claude.ai Drops SSR, Embraces Vite—Performance Jumps 65%
Anthropic engineers replaced SSR with a Vite‑based static stack for Claude.ai, deploying to edge workers and achieving a 65% reduction in p75 TTFB and a 50% faster prompt display, while illustrating why SPA‑oriented AI chat apps may forgo server‑side rendering.
On March 20, Anthropic engineer Felix Rieseberg announced that Claude.ai’s front‑end architecture was rebuilt, moving from server‑side rendering (SSR) to a static solution built with Vite and TanStack Router and deployed on edge workers.
Performance impact: the p75 time‑to‑first‑byte (TTFB) dropped by 65%, prompt‑display speed improved by 50%, and page navigation felt noticeably faster.
The change was motivated by Claude.ai’s SPA‑like interaction model: users maintain a long‑lived chat session with WebSocket streams, making SSR’s SEO and first‑paint benefits irrelevant while adding unnecessary server‑side compute.
The new stack consists of:
Vite : build tool and dev server that replaces the previous SSR framework (likely Next.js).
TanStack Router : client‑side router offering type safety, file‑based routes, and code‑splitting.
Edge Workers : static assets served directly from edge nodes, eliminating the need for rendering servers.
This mirrors a broader “return to SPA” trend; after a wave of SSR popularity with frameworks like Next.js and Nuxt, many teams now reconsider whether SSR is truly required for their product.
Static deployment reduces TTFB because edge nodes return pre‑compiled HTML without server‑side rendering overhead. A 65% reduction means the “white‑screen” time for three‑quarters of users is cut by roughly two‑thirds.
Community reaction was strong: Vue creator Evan You retweeted the news, calling it a fitting win for Vite. Developers noted the case as a live benchmark, and Vite’s recent momentum is highlighted by Cloudflare’s vinext project, which rebuilt Next.js’s API layer on Vite, achieving a 4.4× faster build and 57% smaller bundles.
However, some voices caution that Next.js can also output static pages, so the performance issues may stem from using the wrong mode rather than the framework itself.
Felix concluded that the work is ongoing, emphasizing a culture of incremental improvement to make Claude.ai better each day.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
