On page 16 of his book on symmetric spectra, Stefan Schwede defines the $n$-th Eilenberg-Mac Lane space $(HA)_n$ of an abelian group $A$ by means of a construction called the reduced $A$-linearization. Given a pointed space $K$, the reduced $A$-linearization $A[K]$ consists of all linear combinations of points in $K$ with coefficients in $A$, where all multiples of the basepoint are identified with $0$. This is then topologized via the obvious surjection coming from $\bigsqcup_{n\geq 0} A^n \times K^n$.
Suppose that $x =\sum_i a_i f_i$ is a singular $n$-chain in $K$ with coefficients in $A$. This then defines a map $\hat{x} \colon \Delta^n \to A[K]$ by adding the $f_i$ pointwise and multiplying with the coefficients. Schwede claims that if $x$ is a cycle, the map $\hat{x}$ sends $\partial \Delta^n$ to $0$. This would then give a map $\widetilde{H}_*(K; A) \to \pi_*(A[K])$. However, I don't see why $\hat{x}|_{\partial \Delta^n} = 0$. Suppose for instance that $A = \mathbf{Z}/2$ and $K = \Delta^2$. Then the sum of the three edges gives a 1-cycle $x$ on $\Delta^2$, hence a map $\hat{x} \colon \Delta^1 \to \mathbf{Z}/2[\Delta^2]$. But $\hat{x}(0)$ is the sum of the three vertices of $\Delta^2$, which is not $0$. What am I doing wrong?