Search for a command to run...
Articles tagged with #prompt-caching
Originally published at recca0120.github.io [The first post]({{< ref "/post/claude-code-cache-ttl-audit" >}}) scanned 95 days of logs and found sub-agent cache TTL silently dropped to 5m. [The second]({{< ref "/post/claude-code-cache-ttl-17-days" >}}...
Originally published at recca0120.github.io This past week, chasing a vague "quota burns faster lately" feeling, I scanned three months of my own Claude Code logs. ~\$127K equivalent cost, 127K turns, four models, hundreds of sessions. The uncomforta...
Originally published at recca0120.github.io [Two weeks ago]({{< ref "/post/claude-code-cache-ttl-audit" >}}) I scanned 95 days of Claude Code logs and found that since 4/9, sub-agents had been 100% downgraded to 5m TTL — 5 consecutive days, 4,840 API...
Originally published at recca0120.github.io The previous post covered prompt caching cost mechanics. While researching it I bumped into a dramatic controversy — in March 2026 Anthropic silently changed Claude Code's cache TTL from 1 hour back to 5 mi...
Originally published at recca0120.github.io A common intuition among developers: Claude Code sessions get expensive over time. Context keeps accumulating, every turn resends the entire history, and token costs add up linearly. The obvious conclusion:...