Linear's speed dissected in a technical breakdown

Linear, the SaaS project‑management tool, delivers sub‑second UI updates by pairing GraphQL‑driven data fetching with aggressive caching and incremental loading. The technical breakdown on performance.dev details each layer of the stack and quantifies its impact on latency [performance.dev].

── What shipped ──

• Caching layer – Linear stores query results in a client‑side cache, cutting repeat network traffic by roughly 40 % compared with a naïve fetch approach.
• Incremental data loading – Instead of pulling full tables on every view change, the client requests only the rows that have changed, shrinking payloads to under 10 KB for most list updates.
• Optimized rendering – The UI runs on React’s concurrent mode, allowing the browser to pause and resume rendering while data arrives, which keeps frame times under 16 ms.
• Architecture – A clear separation between the data layer (GraphQL server) and the presentation layer (React components) lets each side be tuned independently, reducing overall bundle size and start‑up time.

These choices collectively keep the main thread idle for less than 200 ms on a typical task board load, according to the breakdown’s measurements [performance.dev].

── Why it matters ──

The pattern shows that latency gains are achievable without wholesale rewrites: caching, incremental loading and a modern rendering engine together deliver a perceptible speed boost. Engineers can benchmark their own stacks against Linear’s numbers to identify similar bottlenecks and prioritize fixes.

Poll: What’s the most critical factor in optimizing web‑app performance?