The reward functions in the problems that they proposed alphaevolve are easy. The reward funtions of at least 50% of maths are not. You can say that validating

The reward functions in the problems that they proposed alphaevolve are easy. The reward funtions of at least 50% of maths are not. You can say that validating if a proof is correct is a straightforward reward, but the size of interesting theorems over the space of all theorems is very small. And also what does "interesting" could even mean?