An attack is a reason to believe that you can solve the problem. I have no idea how'd I go about solving P=NP, but I did have some thoughts on provable security against transient execution attacks. Which is why I work on the latter but not the former.