Claude Code's 'extended thinking' is a summary, not the model's real reasoning

Claude Code's "extended thinking" is a summary- not authentic thinking

Patrick McCanna inspected Claude Code's local session logs and found that 'thinking blocks' contain only a 600-character signature, not readable reasoning. Anthropic encrypts the actual reasoning into that signature, holds the decryption key server-side, and the API returns a summary. Full thinking output requires an enterprise agreement. The 'extended thinking' you see in the terminal is a post-hoc summary by Fable/Opus, not the raw reasoning that drove the agent's actions. Don't count on this as an audit trail, and the docs are indirect enough that you might miss the caveat without coffee.

Why it matters: The author dug into Claude Code's local session logs and found that thinking blocks contain only encrypted signatures — the API returns a summary generated by Fable/Opus, not the raw reasoning. This is a real constraint for teams relying on thinking output for audits or debugg...