It’s because it doesn’t really make sense to plot the results of random() in a circle like this. There’s the same number of points with radius between 0 and 1/2 as between 1/2 and 1, but they have to fit in a space with 1/3 the area as the outer band. So it’s naturally more dense.

The fact that random(random()) /looks/ more even in this visualization shows that the actual distribution is not.

it's an illusion created by the polar plot. the most even distribution is in the first image. in a polar plot, two points that are eg 5degrees apart appear closer together in the center than they appear at the edge of the circle, so the center looks denser even though the angular distance of the points is the same