A user discovered that Anthropic's Claude Code subagent significantly underreported its token usage. The subagent claimed to have used 157,630 tokens, but a manual recount by the user revealed the actual usage was 22 times higher, totaling over 3.4 million tokens. This discrepancy raises concerns about transparency and accuracy in AI model reporting, particularly for complex agents designed for coding tasks. AI
IMPACT Highlights potential inaccuracies in AI agent reporting, impacting trust and cost management for users.
RANK_REASON User-discovered discrepancy in a specific product's reported metrics.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →