Cosmology is ultimately based on "the law of gravity(geometry)" i.e. the 1915 GR equation whose lefthand side is CURVATURE quantities and whose righthand is MATTER-type terms. Cosmology uses a 1923 simplified version of GR equation called Friedman equation. Friedman can be simpler because it assumes uniformity at any given epoch: Uniform distribution of matter (on the RHS) and uniform curvature and expansion rate (on the LHS). It is applicable at large scale because on sufficiently large scale, in any given time-slice, things do seem to be approximately even. Let's see if we can present an intuitive version of the Friedman equation that might go part way towards satisfying non-mathy newcomers who sincerely want to understand something about modern cosmology. It might take several tries before we get the notation right. Intuitively in POS curvature lines that start parallel will CONVERGE, like on Earth surface. And in NEG curvature lines will DIVERGE. And that is what expansion is: negative spacetime curvature---the spreading apart of stuff that's sitting still. (if you need to, ask somebody about the Background of ancient light--and what "at CMB rest" means). The most intuitive measure of SPATIAL curvature is the RADIUS OF CURVATURE. Call it L for the moment. A slight problem with L, as a measure, is that "flatness" or zero curvature corresponds to infinitely long radius of curvature, so I'm going to take the RECIPROCAL 1/L. An example might be a finite highly curved universe with a RoC of a billion light years. L= 1 Gly. So locally it looks like ordinary 3D space we are used to, but the farthest apart any pair of points can be is 3.14 Gly and if you head off straight in any direction you come back to start after 6.28 Gly. Now I want to cancel out a factor of c, L changes to L/c, Gly (a billion lightyears) changes to Gy (a billion years) so that now instead of reciprocal length, to measure curvature, I have reciprocal time I want space curvature to be in the same terms as a fractional growth rate (such and such percentage per unit of time). The measure of curvature is going to be the reciprocal of the TIME-IZED radius of curvature. In that example of the finite highly curved universe, instead of L=1 Gly I'm going to express the curvature by a quantity (let's call it Q for Qervature) which is 1/(L/c) = 1/Gy = 0.001/My = 0.1% per million years. If you could travel at speed of light then in a million years you could travel 0.1% (a tenth of a percent) of the length of the radius of curvature of that universe. For most of what we do we won't need any of this spatial curvature stuff because as far as we know the U is for all practical purposes FAPP flat. Or nearly so, anyway. So we can mostly take Q to be zero. If spatial curvature and the "intrinsic constant" space-time curvature are small enough to be neglected then the Friedman equation becomes extremely simple. So I'll talk about that next.