Comment by isolli

Why is supervising one's agents to be as efficient as possible a problem for Anthropic?