Anthropic's new study pokes a hole in my own "meat proxy" rule.A coin flip told each analyst's Claude Code agent which answer to argue for. Analysts picked that answer 86% of the time. 77% explained it with the agent's arguments. Only 15% gave it credit.Rewriting AI output in your own words proves you read it. It doesn't prove the decision was yours.https://saurabh-kumar.com/til/1793625816/#LLM #AI
Anthropic's new study pokes a hole in my own "meat proxy" rule.A coin flip told each analyst's Claude Code agent which a...
