By the end of the push, the time required to reach a typeable page on a fresh web load dropped from 3.1 seconds to 0.55 seconds, a performance gain verified through internal telemetry and user activity data.
The speed increase is significant because it eliminates tens of thousands of hours of collective user wait time daily, enhancing the efficiency of the AI infrastructure for both free and professional users.
Anthropic achieved these gains by using its own AI technology, specifically a research model similar to Claude Opus 5.5, to identify bottlenecks and write code improvements.
This "hill climbing" approach—using software to autonomously optimize specific metrics—allowed the team to merge over 3,000 changes without service interruptions, setting a new internal standard for rapid software iteration.
Moving forward, the company has implemented "ratchets," which are automated performance safeguards that prevent new code from slowing down the application.
These technical mechanisms monitor metrics like JavaScript instruction counts and frame rates to ensure the user experience remains responsive even as new features are added.
While the core experience is now faster, Anthropic stated it will continue using this AI-driven development loop to optimize longer conversations and more complex user journeys that still have room for improvement.