Here is a question that sounds too simple to be deep. You carry something on a round trip — out and back to where you started — and you are careful never to twist it, turn it, or fiddle with it along the way. When you get home, is it the way it left? On a flat tabletop: obviously yes. On the surface of the Earth: no. It comes back rotated, by an angle you can predict exactly, even though you did everything right. This section is three experiments with that fact: the round trip itself, a surface that hides all of its curvature in a single point, and two roads that honestly deliver two different answers.
The rule of the game is honesty: you hold an arrow flat against the surface you are walking on, and at every step you keep it pointing the way it was pointing — no deliberate turning, ever. If the path bends, you do not bend the arrow with it; you just keep it as straight as the surface allows. This careful no-twisting carry has a name: parallel transport. It is the only fair way to compare a direction here with a direction there: walk it over.
Below, the same trip runs on two surfaces at once. On the left, a flat patch: walk the triangle, and the arrow comes home exactly as it left — always, no matter how big the triangle. On the right, a globe: walk down to the equator, along it, and back up to the pole. Watch the arrow closely on each leg — it is never turned, it only rides. And it comes home rotated anyway. Widen the triangle and the rotation grows: the mismatch angle equals the area the walk enclosed (in radians, on a globe of radius one). The surface's curvature, which you cannot see from any single point, is sitting right there in the returned arrow.
Two things to hold onto from the figure. First, the rotation is not an error — there is no more-careful way of carrying that makes it go away. It is a property of the surface, not of your technique. Second, it is proportional to enclosed area: tiny loops pick up almost nothing, big loops pick up a lot. Curvature is invisible locally and unmistakable around a loop. That is why the loop — not the point — is the natural instrument for measuring what a space is really like.
Now a stranger surface. Take a flat disk of paper, cut a wedge out of it, and glue the two cut edges together: a cone. Here is the point of the experiment — every part of that cone is still flat paper. You did not stretch it anywhere; an ant standing anywhere on the slope could run every local test it likes and find nothing but ordinary flat geometry. All of the missing wedge got concentrated into one place: the tip.
So walk a loop around the tip, carrying your arrow honestly. On the left below, the cone is shown unrolled — flat, with the glued seam marked — and there, transport is trivial: the arrow simply holds its direction, because the paper is flat. But the walk ends at the seam, and the seam is a lie the flat picture tells: crossing it means re-entering at the other cut edge, and the arrow comes back rotated by exactly the wedge you removed. On the right, the same walk on the assembled cone. Slide the wedge and watch the returned arrow track it, degree for degree.
Hold onto this one — it is load-bearing for the whole site. A surface that is flat almost everywhere, with its curvature concentrated into point defects that only loops can detect, is exactly the picture behind the vortices of The Physics of Mind §05: the phase field is smooth nearly everywhere, and the interesting physics — the defects whose binding and unbinding is the phase transition — lives at cone-tips like this one. When that chapter says a vortex has a winding that "cannot be smoothed away," it is saying: there is a cone-tip here, and loops around it know.
One more consequence, and it is the one that makes “carrying” genuinely different from “copying”. On a flat surface, if two couriers set out from the same point with identical arrows and deliver them to the same destination by different routes, the deliveries agree — always. Direction can be copied across flat space, and the route is irrelevant. On a curved surface, try it:
This is why, on a curved surface, there is no such thing as “the same direction, over there.” Direction is not a global fact you can copy; it is a local fact you must carry, and what arrives depends on the road. The disagreement between roads and the rotation around loops are the same phenomenon — a loop is just two roads glued back to back — and both measure the same thing: the curvature caught in between.
Next question: to even say “the arrow points at 37°” you needed a reference direction at every place the arrow visited — and who chose those? Nobody, it turns out, and everybody. That is gauge, and it is where the bookkeeping becomes physics.