The referenced sequence records numbers whose largest prime factor falls short of the square root of the number. The graph of the sequence makes it clear that these numbers have a well-defined asymptotic density. Following a citation to HAKMEM 29 (Schroeppel), we find a claim equivalent to saying that this density is 1 - ln 2 ~ 0.30685 But empirically, the density looks considerably smaller, more like 0.27. The problem can't be whether or not to include exact squares of primes, because those have asymptotic density 0. We have one more reported reference, to Exercise 28 on p. 26 of Tenenbaum and Wu, Théorie analytique et probabiliste des nombres, solution on p. 34. But I don't have a copy, and my French isn't up to the task of understanding their solution. Can somebody check it and compare it with Schroeppel's result (which, typically for HAKMEM, is stated without proof)? Rich, do you remember your approach?