Detailed Analysis
A Claude Max subscriber reports a sudden and severe degradation in service reliability beginning around Friday, May 8, 2026, marking a sharp contrast to what the user describes as "amazing" performance over the preceding two months of consistent use. The core symptoms include repeated API timeout errors during routine tasks, persistent stalling that requires constant manual intervention to restart processes, and an inability to complete straightforward assignments — such as drafting website copy — without continuous user supervision. The user reports the problems occurring across both Claude's standard interface and its Code environment, suggesting the issue is not isolated to a single product surface but likely rooted in a backend infrastructure problem affecting the platform broadly.
The pattern of behavior described — working normally for a few minutes before stalling again — points toward intermittent server-side failures rather than a complete outage, which can be more frustrating for end users than total unavailability. Intermittent failures are notoriously difficult to diagnose and work around, as they create false confidence that the system is functional before breaking down again. The user's description of "simple tasks now taking hours" quantifies the productivity impact in concrete terms and illustrates how even partial API reliability failures can effectively render a paid subscription non-functional for professional use cases.
The incident raises important questions about service-level expectations for premium AI subscriptions. Claude Max is Anthropic's highest consumer tier, and users at that level typically have elevated expectations for reliability and uptime. The user's framing — noting the experience "has basically become unusable" — reflects a growing sensitivity among power users to infrastructure stability as a core product feature, not merely a secondary concern behind capability benchmarks. When AI tools are integrated into professional workflows, even brief periods of unreliability can cascade into significant operational disruption, as routine tasks that once took minutes can balloon in time cost.
This type of reported degradation fits into a broader pattern observable across the AI industry, where rapid scaling of user bases and computational demand periodically strains infrastructure in ways that produce inconsistent performance. As Anthropic has expanded Claude's subscriber base and introduced more capable models with higher computational requirements, the backend systems must continuously scale to match demand. Periods of unexpected traffic spikes, model deployment transitions, or infrastructure maintenance can create precisely the kind of intermittent timeout behavior described here. The user's appeal to the community for shared experiences and fixes also reflects a broader phenomenon: when official support channels are slow or opaque, technically sophisticated users increasingly turn to peer communities like Reddit as first-response diagnostic resources, effectively crowdsourcing incident detection and triage that might otherwise surface more slowly through formal support pipelines.
Read original article →