$X_{Ab}$ is the Quillenization of a path-connected space $X$ if $X_{Ab}$ has abelian fundamental group, and there exists a continuous map $X\rightarrow X_{Ab}$ inducing an isomorphism $H_n(X)\rightarrow H_n(X_{Ab})$ for all $n$.
Let $\pi=\pi_1(X)$. The claim that I don't understand is that $X_{Ab}$ exists if and only if $[\pi,\pi]=[\pi,[\pi, \pi]]$.
It seemed at first to me that the proper approach to seeing this would use the result that $f:S^1\rightarrow X$ is in the kernel of the Hurewicz map iff it can be extended to a map of a "disk with handles", which easy to see since generators of $[\pi,\pi]$ form 1-skeletons of closed oriented surfaces. Now, $[\pi,[\pi,\pi]]\subseteq[\pi,\pi]$ in general, and classification of surfaces gives that elements of $[\pi,[\pi,\pi]]$ are 1-skeletons of those oriented surfaces with an odd number of handles, if I've done my counting correctly (which is a tenuous claim at best...)
Trouble is, I don't see how this implies the statement at all. If $f:X\rightarrow X_{Ab}$ is the obvious map which collapses loops in the 1-skeleton of $X$ homotopic to elements of $[\pi,\pi]$, then it clearly induces an isomorphism on homology for $n>2$, and for $n=2$, it doesn't affect $\ker\partial_2$, since the only really troublesome case is $H_1$. Why shouldn't this $f$ induce an isomorphism there, too? It's just sending things to 0 which would get sent to 0 anyway by the Hurewicz map.
I'm not looking for an answer, but if someone could point out at what point I'm clearly thinking about this incorrectly, I'd greatly appreciate it.