Carlson, K. OpenAI: Toward Mechanistic Interpretability (MI). SI 2025, 2.