Anthropic says it made Claude.ai and the Claude desktop app's core UX ~3x faster in August using an internal model to ship improvements "in a two-week sprint" (Anthropic)

In August, Anthropic's Claude.ai and Claude desktop app experienced a significant speed boost, according to findings from the company's internal research model. The improvements, made within a two-week sprint, saw the core user experience of the platforms accelerate about three times faster. Users had previously reported the platforms as slow, and the data validated these concerns.

The sprint focused on four user journeys that account for 95% of activity. The time to type a page on a fresh load of Claude.ai dropped from 3.1 seconds to 0.55 seconds; starting a new Claude Code session went from 0.8 seconds to 0.3 seconds; and loading a Claude Cowork cloud session reduced from 2.6 seconds to 0.73 seconds. In total, this optimization is estimated to have saved tens of thousands of user-hours per day.

The team used Claude Tag (beta) to run an internal research model, comparable to Opus 5.5, to identify bottlenecks, build benchmarks, and ship improvements. More than three thousand changes were merged without any customer-facing incidents or rollbacks. The team created a Slack channel named "Claude," with the mission to facilitate performance-related tasks.

Claude was tasked with monitoring deploys for performance regressions, assessing telemetry accuracy, maintaining observability dashboards, implementing solutions for observed issues, proposing performance projects, and communicating with human teammates. The sprint resulted in the optimization of twelve out of thirteen identified targets by day three.

The team applied a strategy of setting goals, making tradeoffs, and approving every change, leading to a successful sprint with no incidents or rollbacks.