Economic policies should be evaluated based on their real-world impact.
The Langlands program is not a theory but a web of conjectures and theorems that, if fully established, would unite number theory and representation theory.
The best way to predict the future is to create it.