Of course he'll say that. But even removing the potential of lying on his par6, I'm sure his reports are telling him things like: This latest model we trained

Of course he'll say that. But even removing the potential of lying on his par6, I'm sure his reports are telling him things like: This latest model we trained on X% more data by integrating novel training data. We now see user conversion funnels working better. User feedback in closed betas indicate that 89.6% feel like answers are "generally correct most of the time". Etc etc.