Claude Code users hitting usage limits 'way faster than expected'
Why it matters: Disruptions to Claude Code can impede software developers' daily workflows and productivity.
- Anthropic is investigating an issue where Claude Code users are hitting usage limits 'way faster than expected,' making it a 'top priority' to resolve.
- Customers are frustrated, with some reporting that paid accounts deplete tokens faster than free accounts, and a single session can drain a daily budget in minutes.
- Anthropic recently introduced peak-hour throttling on Claude, which consumes tokens more quickly during high demand.
- Anthropic accidentally released 500,000 lines of Claude Code's internal source code on GitHub due to 'human error,' not a security breach, with no sensitive customer data exposed.
- The leaked source code reveals details about Anthropic's future plans, including a 'Tamagotchi-style pet' and an 'always-on agent,' as reported by The Verge AI and Ars Technica.
- Independent developers had already partially reverse-engineered Claude Code's source code, and an earlier version was leaked in February 2025.
Users of Anthropic's AI coding tool, Claude Code, are hitting usage limits much faster than expected, prompting the company to prioritize a fix amid customer frustration over opaque token consumption and recent peak-hour throttling. This issue is compounded by a recent accidental leak of Claude Code's internal source code, revealing insights into Anthropic's future AI plans, including a 'Tamagotchi-style pet' and an 'always-on agent,' according to The Verge AI and Ars Technica.
