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

[jessegrabowski]

One could use the mean if the optimizer was already on the graph informing that mean, but not otherwise