Its greatest strengths might be within its ability to evaluate different approaches, essentially performing simulations and recodes (not random walks), many times faster that a human could test and review code. Given enough processing power, it could cover far more cases in far less time than crews of test engineers and coders.