Burned $6,744 on Claude Code Sessions—97% Went to Cache Reads, Not Code
Staring at my terminal last night, parsing 187 Claude Code sessions that torched $6,744 in tokens. Turns out, almost everything's just Claude re-reading its own homework.
⚡ Key Takeaways
- 97% of Claude Code tokens are cache reads, making real control tiny at 2.8%.
- Fixes like /compact and grep over Agent slashed anomalies instantly.
- Tool like ccwhy reveals why—open source, offline, game-changer for AI bills.
Worth sharing?
Get the best Developer Tools stories of the week in your inbox — no noise, no spam.
Originally reported by dev.to
Related Stories
AI Dev Tools
17-Point AI Performance Gap from Bad Instructions — And the Tool Fixing It
Open Source
X's Open Algorithm Powers This Free Chrome Extension That Predicts Tweet Reach Live
Open Source
12-Year-Old Drops 2KB PermzPlus Bomb on CASL's Bloat
Open Source