That surprised me since I saw the "generalization to digits beyond the first" in the article and it's clear that powers of 2 or 5 clearly don't have a Benford-s

That surprised me since I saw the "generalization to digits beyond the first" in the article and it's clear that powers of 2 or 5 clearly don't have a Benford-style distribution of their last digits. However, that was a misunderstanding on my part, since that generalization says something about the first n digits, and any fixed n will eventually not include the last digit of the number.