“And the fundamental point of all these massively parallel experiments is the same: when a problem reaches a certain level of complexity, formal theory won’t get you nearly as far as an incredibly rapid, systematic process of trial and error.”