CATCH #782
DETECTED AUG 17, 2026 21:13 UTC
THE CATCHscore 6/10
EXHIBIT 782 · URGENT
Aug 17, 2026 21:13 UTC
web app
https://platform.claude.com
Anthropic's Claude.com console frontend code includes new strings hinting at a breakdown of output tokens by category, specifically detailing internal reasoning tokens for observability.
READY TO POST:
looks like claude's platform might be getting more detailed token usage breakdowns, including internal reasoning. could be super helpful for cost analysis!
anthropic's frontend code points to new observability for output tokens, separating out internal reasoning. curious to see how that impacts prompt engineering.
EVIDENCE — the receipts
⚠️ FROM WEB-APP BUNDLE — may include vendored-dependency strings, verify before posting.
Anthropic Console (https://platform.claude.com) deployed new frontend code [aca8e3c38b49193a -> 292aeed8215fe7db]: +1314 strings, -1352 strings, 3 interesting
[feature_flag] (2)
+ Breakdown of output tokens by category.
`output_tokens` remains the inclusive, authoritative total used for billing.
This object provides a read-only decomposition for observability — for example,
how many of the billed output tokens were spent on internal reasoning that may
have been summarized be
+ Number of output tokens the model generated as internal reasoning, including
the thinking-block delimiter tokens.
Reflects the raw reasoning the model produced, not the (possibly shorter)
summarized thinking text returned in the response body. Computed by
re-tokenizing the raw reasoning text, so it
[endpoint] (1)
+ https://support.claude.com/en/
caught by the 20-minute sweep ·
the live wire · all receipts