One-shot debugging on write
1.6M tokens vs typical 215.4K
4m session running commands and testing. 70K tokens across 1 prompts using claude-sonnet-5-5.
Derived from this session's token and cost data. Not shown on the feed.
Comments