I could do a tiny bit more with my methods, and, wham! Steve found the right method. It was in a paper of Walfisz, and it finished the job; Dirichlet had less than half solved the problem… probably not losing any sleep over it… and Isbell and Schanuel finished it. That’s in Proc. AMS 60 (1976). 65-67, if any of your readers want the theorem. It’s a very nice theorem, but there’s a little story about it that must be told. At the next annual meeting of the AMS, the first evening I go into a restaurant with a couple of other guys, and there is Paul Erdös coming in with a couple of other guys. (Maybe gals, I don’t recall). So I say “Paul! let $j = o(n)$, etc. etc.” I get through the hypothesis, and I pause for breath. And he tells me the conclusion. So I say, “Oh, my God. Is that your theorem?” “No,” he says. “It’s a nice theorem. I never heard about it; but if there is a theorem, and that is the hypothesis, then this must be the conclusion.