Hornbein said:
I've been told, that if the curvature is non-zero that it increases over time.
This is referring to a particular definition of
spatial curvature (and in a particular type of model, see below), not
spacetime curvature. As
@PAllen posted, the
spacetime curvature of any expanding FRW solution will decrease with time.
In matter and radiation dominated FRW spacetimes, if the density parameter ##\Omega## (the ratio of actual density to critical density) is not exactly ##1##, then it will diverge from ##1## over time. In models with zero cosmological constant, ##\Omega - 1## can also be treated as a (sort of) spatial curvature parameter; if it is zero (i.e., ##\Omega = 1##, density exactly critical), the universe is spatially flat; positive ##\Omega - 1## means positive spatial curvature (closed universe), and negative ##\Omega - 1## means negative spatial curvature (open universe). A universe with positive ##\Omega - 1## will have increasing ##\Omega - 1## over time, whicn can (sort of) be thought of as increasingly positive spatial curvature, and a universe with negative ##\Omega - 1## will have decreasing ##\Omega - 1## over time, which can (sort of) be thought of as increasingly negative spatial curvature.
(Note that I put in the qualifier "sort of" each time I mentioned spatial curvature in the previous paragraph. That's because ##\Omega - 1## is
not the actual geometric quantity that would normally be referred to as the "spatial curvature" of a spacelike slice of constant FRW coordinate time in these spacetimes. Sources that talk about it as related to "spatial curvature" without any qualification are being sloppy.)
The reason cosmologists are interested in all this is that, for most of our universe's history, it was either matter or radiation dominated, and yet our current universe's ##\Omega## is extremely close to ##1##. This requires extremely sensitive fine-tuning of initial conditions, which is known as the "flatness problem":
https://en.wikipedia.org/wiki/Flatness_problem