This model then only makes sense for computations that are easier to verify a solution for than to solve (or where both are trivial in terms of computation), presumably?