Debugged ai-test-harness
336.4K tokens vs typical 22.1K
9m session running commands and testing. 336K tokens across 4 prompts using claude-opus-4-8[1m].
Derived from this session's token and cost data. Not shown on the feed.