Claude.ai users report a critical auto-compact feature has failed since a major outage on January 14. The bug, tracked in a GitHub issue, causes messages to silently bounce back or show a 'limit reached' error, forcing manual intervention to continue conversations. Anthropic marked it as fixed on January 15, but the problem persists.

This regression breaks a core workflow for long-form projects, where auto-compact should seamlessly manage the 200k token context window. Without it, users must restart chats every 30-45 minutes, losing productivity and context. The silent failure offers no error messages, making the tool unreliable for sustained work despite a $200 monthly subscription.

Users have contacted Anthropic engineers for updates, but there's no official acknowledgment or timeline for a fix. The issue highlights the fragility of core AI assistant features post-outage and the need for transparent communication. Until resolved, paying customers are considering alternatives, undermining trust in Anthropic's platform stability.