Anthropic Silently Downgraded Cache TTL from 1 Hour to 5 Minutes — Developer Documents $2,500+ in Waste

Hacker News·medium signal

A Claude Code user analyzed 119,866 API calls across two machines and documented that Anthropic shifted prompt cache TTL from 1 hour to 5 minutes on March 6, 2026. The cost impact: $949 overpayment on Sonnet 4.6 and $1,582 on Opus 4.6 (17.1% waste each). Cache writes cost 12.5x more than reads, so any coding pause >5 minutes forces expensive re-creation. Anthropic's Jarred Sumner confirmed it was intentional, not a regression, and closed the issue as NOT_PLANNED. 107 points and 81 comments on HN.