/tech

Claude’s web and desktop apps reportedly got faster through a two-week optimization sprint

RuntimeWire, citing Anthropic, reports a 3.1x geometric-mean improvement across 13 performance measurements—not a threefold speedup for every interaction.

TagsAnthropic

5 Sources, ,

Anthropic Engineering card for its report on making claude.ai faster.

TLDR

RuntimeWire reports that Anthropic’s August 2026 sprint used Claude to find bottlenecks, build benchmarks, ship fixes and monitor deployments. Anthropic says engineers set goals, weighed tradeoffs and approved every change. At the 75th percentile, fresh-load time until users could type on claude.ai fell from 3.1 seconds to 0.55 seconds, according to the company. Anthropic reports merging more than 3,000 changes without a customer-facing incident or rollback, using automated review, tests and staged rollouts. The team says very long conversations, other user journeys and 95th-percentile performance still have room to improve.

Featured Source

Anthropic says a two-week performance sprint in August made the core experience in claude.ai and the Claude desktop app about three times faster. The headline figure is a geometric mean across 13 measurements, so it does not mean every action became exactly three times faster.

Anthropic's engineering account says the team focused on four journeys that represent 95% of user activity: launching the app, starting a conversation, loading an existing conversation and sending a message. At the 75th percentile, the time from a fresh load to a typeable claude.ai page fell from 3.1 seconds to 0.55 seconds. Starting a new Claude Code session dropped from 0.8 seconds to 0.3 seconds, while loading a Claude Cowork cloud session went from 2.6 seconds to 0.73 seconds.

Claude measured, proposed and monitored the changes

The team used an internal research model called Claude Tag to analyze telemetry, build benchmarks, find bottlenecks, propose fixes and watch deployments. Engineers remained responsible for setting goals, weighing tradeoffs and approving each change. Anthropic says more than 3,000 changes merged during the sprint without a customer-facing incident or rollback.

Some improvements were architectural. The team added a static composer so users could type while React initialized, precompiled a V8 code cache for the desktop shell, kept the composer mounted between conversations and reduced sidebar re-renders. It also built deterministic benchmarks around instruction counts, React commits, layout changes and other work that could be enforced in continuous integration.

Anthropic says the apps are now about three times faster than they were in early August, while noting that the 95th percentile, very long conversations and other user journeys still have room to improve. Those are company measurements rather than independent performance tests.

TLDR

Claude measured, proposed and monitored the changes

Claude’s web and desktop apps reportedly got faster through a two-week optimization sprint

TLDR

Claude measured, proposed and monitored the changes

Claude’s web and desktop apps reportedly got faster through a two-week optimization sprint

7 Sources

Combined views

Claude measured, proposed and monitored the changes

Combined views

Sentiment

Summary

TLDR

Claude measured, proposed and monitored the changes

Sentiment

Summary

7 Sources

Sentiment

Summary