Anthropic reduced the Claude Code prompt cache TTL from one hour back to five minutes in early March, sparking user complaints about faster quota exhaustion. Anthropic's Jarred Sumner defended the change, arguing it lowers costs for one-shot requests, while Claude Code creator Boris Cherny acknowledged that cache misses on the one-million-token context window are expensive and said Anthropic is investigating a 400K-token default. Users report that Pro subscribers ($20/month) may get as few as two prompts in five hours, and several bugs in caching code have been flagged. Beyond caching, some developers believe Anthropic's quotas are simply buying less processing time than before, with reports of degraded model performance and faster session exhaustion since late March.
5.4K Impressions1 Comment