How Claude Helped Make claude.ai Three Times Faster in Two Weeks

Summary

Anthropic says a two-week sprint made the core user experience of claude.ai and the Claude desktop app about three times faster. The team focused on four journeys representing 95% of user activity: launching the app, starting a conversation, loading an existing conversation, and sending a message. At the 75th percentile, time to a typeable page on a fresh web load fell from 3.1 seconds to 0.55 seconds, starting a Claude Code session fell from 0.8 to 0.3 seconds, and loading a Claude Cowork cloud session fell from 2.6 to 0.73 seconds. The team used Claude Tag, an internal research model described as roughly comparable to Opus 5.5, to inspect Datadog data, build benchmarks, propose code changes, and monitor deployments. Claude helped merge more than 3,000 changes without a customer-facing incident or rollback, while humans set goals, approved changes, and judged user-visible tradeoffs. The process relied on deterministic measurements such as instruction counts, React commits, style recalculations, DOM mutations, and controlled 120 Hz frame timing, discarding benchmarks that did not correlate with wall-clock improvements. Examples included a static HTML composer, session prefetching, fewer sidebar rerenders, and targeted fixes to streaming and syntax highlighting. More than 150 optimization threads ran concurrently, supported by human review, tests, feature flags, incremental rollouts, and ratcheting CI guardrails. In one streaming project, total main-thread blocking fell from about 750 milliseconds to 200, with 120 frames per second sustained on a 120 Hz MacBook. The authors say the 95th percentile, very long conversations, and other journeys still need work.