“OpenAI's Superalignment team is not focused on interpretability work for smaller or open-source models because its comparative advantage is working with the largest models.”