i would expect it to make great product decisions for it's own model, and perhaps other LLMs. for humans, why would you assume it would?