Anthropic says Claude Code quotas drain faster
Anthropic’s Claude Code coding assistant is under scrutiny after users reported their usage quotas being exhausted far sooner than normal, sometimes in minutes rather than hours. The Register reports that Anthropic has acknowledged the issue and says it is actively investigating, while users point to a mix of peak-hour quota adjustments, the end of a temporary usage promotion, and possible prompt-caching bugs that may be inflating token consumption. The result is a messy picture for paid subscribers, especially developers running Claude Code in longer or automated workflows.
The core problem is not just “limits exist,” but that the metering feels unpredictable enough to break trust with power users.
- –This is a negative operational update for Claude Code, not a launch or feature win.
- –Anthropic’s public explanation suggests demand management, but user reports indicate something more severe or at least poorly communicated.
- –The most likely audience impact is on Pro and Max users doing long-context or background agent work, where hidden consumption spikes are most painful.
- –If the prompt-caching bug angle is real, this is a reliability issue as much as a pricing issue.
- –The story reinforces a broader market problem: flat-rate AI subscriptions are getting squeezed by heavy agentic usage.
DISCOVERED
57d ago
2026-03-31
PUBLISHED
58d ago
2026-03-31
RELEVANCE
AUTHOR
samizdis