Once Claude can measure something, it can make it faster (claude.dev)

In an impressive two-week sprint, the team behind claude.ai has successfully tripled the speed of their core user experience, addressing user feedback regarding performance issues. By strategically focusing on the four user journeys that comprised 95% of activity, they reduced loading times significantly—down from 3.1 seconds to 0.55 seconds to a typeable page and from 2.6 seconds to 0.73 seconds for loading a cloud session. This enhancement is expected to save users tens of thousands of hours of waiting daily, showcasing the crucial importance of responsiveness in AI applications. The technical backbone of this improvement involved the use of an internal model, Claude Tag, which effectively identified performance bottlenecks and facilitated numerous real-time optimizations. The channel-based workflow allowed Claude to autonomously inspire new measurement methodologies, enabling a rapid cycle of testing and refinement. With over three thousand code changes deployed without customer-facing issues, the initiative exemplifies the power of measurement-driven development within the AI/ML realm. The commitment to continuous performance tracking created a dynamic feedback loop, empowering Claude to proactively pinpoint areas for enhancement, ultimately transforming performance optimization from a reactive to a proactive process.