I wish for either a proof that ECDLP (Elliptic Curve Discrete Log) is in BPP (Bounded Error Polynomial Time) or a proof that it isn't.
But I'd also accept the genie's statement that neither is provable in ZFC...
I'm in good health, pretty happy, and have no wish for material wealth. But as a curious theoretical computer scientist, this is one of the great questions, closely related to P vs NP. Basically, it asks if all of our current public key infrastructure is safe in the absence of quantum computers (which I don't expect to see in my lifetime).