wire media がリツイート

The recent accidental open-sourcing of the Claude Code CLI via a bundled .map file revealed some incredibly funny and slightly dystopian details about how Anthropic measures user satisfaction.
They are, in fact, tracking your temper tantrums.
If your prompt contains strings like "wtf", "ffs", "this sucks", "piece of shit", "pissed off", or standard profanity, the local CLI catches it before it even hits the LLM.
It silently attaches a boolean flag "is_negative: true" to the tengu_input_prompt telemetry event and ships it off to Anthropic's analytics dashboard. (It also tracks an is_keep_going: true flag for phrases like "continue" or "go on").

rahat@Rahatcodes
Claude Code has a regex that detects "wtf", "ffs", "piece of shit", "fuck you", "this sucks" etc. It doesn't change behavior...it just silently logs is_negative: true to analytics. Anthropic is tracking how often you rage at your AI Do with this information what you will
English


















