The hardest single part of building a software system is deciding precisely what to build.
The Langlands correspondence is not just a theorem but a guiding principle for future discoveries.