Can I make out-of-sample predictions using posterior means? Or how to use scipy.optimize with pymc predictions?

I don’t follow?