Imagine two people standing some distance apart from each other at the equator. They both begin walking in straight-line paths due south. At first, their paths are parallel. But as they move toward the south pole, they begin to drift closer to each other, as though their paths were curving towards each other. When they reach the south pole, they bump into each other. But they were both walking straight forward following the shortest path to the south pole the whole time. The curvature of the surface causes their initially-parallel paths to converge.[1]
On a plane (which has Euclidean geometry), initially-parallel paths never converge.
[1] Don't take this too literally; the real planet Earth is three-dimensional, and its gravity keeps us on the surface. But mathematically, it's possible to describe a curved two-dimensional space without referring to any higher dimensions. When I talk about "the surface of a sphere", that's what I mean -- the surface is the entire 2D space.
Space-time is 4d array: array of framebuffers. You can stretch your mathematical model all day long, but you knowledge must be mapped to reality somehow. In model we have space-time, while in real world we have "physical vaccum" ("something nothing" or "phaccuum", for short). I prefer to name that thing "ether", because I like that word.
In spherical geometry, the equivalent of a straight line is a great circle. There are no parallel great circles. That's why I used the phrase "initially parallel" -- at the starting point, both people's paths are at a 90-degree angle to the great circle connecting their locations.
I didn't want to get into "locally flat" vs. "globally curved" in something that started as an ELI5 thread.
Yes, of course. If we substitute parallel lines with straight lines in spherical geometry and mix 2D and 3D spaces, then our mental model will be nonsensical but cute.
We found no evidence of fourth dimension in the real world, so we cannot map this cute mathemagical model to reality.
Same thing in general relativity: the metric tensor measures the failure of closed loops on each axis to not close perfectly, the way they would in Euclidean space.
Basically even as a small creature on earth you can 'figure out' about the curvature by carefully measuring small-ish loops. The same is true for spacetime, but the loops' deformities are even smaller.
Take a straight line down from the "north pole" of your ball to its equator. Draw another straight line around a quarter of the equator. Draw a third line back to the pole. You've just drawn a triangle with 3 straight lines and the angles add to 270 degrees.
A non straight line is just not the shortest distance between two points on that surface.
Shortest distance between two points is what it is.
If the 'tether' is a gravitational link (meaning that the teather, is a constant pull against the trajectory, regardless of the trajectory, the object will continue to curve around center.