# Talk:99 questions/Solutions/35

### From HaskellWiki

The last change to the code brought its complexity down to the theoretical one of trial division, as it should be, about O(n^1.45) empirically, in number of primes produced; previous versions had it up there in O(n^1.67) or even O(n^1.85), which is what I've started from. WillNess 12:02, 31 May 2011 (UTC)

- this refers to using
`filter (\n-> n==head(factor primes n))`

instead of`filter (null . tail . factor primes)`

. But actually the*external*faster-produced primes list of Q.31 should just be used. WillNess 23:55, 4 June 2011 (UTC)