Because We Have LLMs, We Can and Should Pursue Agentic Interpretability