That’s roughly equivalent, as I understand it, as it returns 2^{53} evenly spaced unique values between 0 and 1. The thing that makes the approach in Output distribution of rand() (thread 1) and the linked blog post clever is that it returns 2^{53} evenly spaced unique values in [0.5, 1), 2^{53} twice-as-closely-spaced unique values in [0.25, 0.5) (with half the probability), and so on…