The Oldest Orientation Problem — Paper 02 of The Geometry of Meaning
Arc ·
Point straight up. Actually do it, if nobody’s watching — the ceiling doesn’t mind. Notice how little thought it required. Of all the questions in this series, which way is up may be the only one your body answers before your mind is consulted. You have known up since before you had words: it is the direction sleep falls away from, the direction thrown things betray, the direction the ground forbids. Ask a hundred people to point straight up and you get a hundred confident arms and no arguments.
The previous paper ended in trouble: two butterflies in two private universes, no shared signposts, no way to compare coordinates that were never issued by a common authority. I promised that the walk toward an answer would begin somewhere unexpected. This is it. It begins with your arm, raised, certain — and with the quiet demolition of that certainty, which our species already survived once, and profited from beyond all imagining.
Because here is the thing about the hundred confident arms: they do not agree. They have never agreed. And it took us most of recorded history to notice.
Stand in Denver and point straight up. At the same instant, have a friend in Perth — western Australia, most of a planet away — do exactly the same thing. Both of you are performing the same act, with the same certainty, described by the same two words. And your arms point in nearly opposite directions through three-dimensional space. The angle between them is about 166 degrees — fourteen degrees short of perfectly opposed. If your two ups were both extended as lines into the sky, they would never meet, and each of you is aiming, roughly, at the soles of the other’s shoes.
Sit with the strangeness before dissolving it. Neither of you is wrong. There is no tribunal that can rule for Denver and against Perth; no experiment either of you can run, locally, that reveals your up to be the deficient one. The word is doing exactly what it has always done, perfectly, in both mouths. And the two perfect performances contradict each other — unless the word never meant what you assumed it meant.
I should confess what this example is, because it is not an illustration I went looking for. It is where this entire research program began.
Two people point straight up and they’re pointing in opposite directions and they are both right. Which means “up” was never a direction. It was a relationship wearing a direction’s clothes. And now I can’t stop asking — how many other words are doing this?
So: up is not a direction. What is it, then, exactly? For that we need the oldest scientific instrument that is also a toy — a rock on a string.
Hang a small weight from a thread and hold it still. The thread draws a line. Every builder since Egypt has trusted that line, and it is worth noticing what an odd act of trust that is: the thread consults nothing visible. No landmark, no star, no authority. It simply settles — and where it settles, we call vertical. Geodesy, the science of measuring the Earth, gives the two ends of that line names: extend it downward and you reach the nadir; extend it upward, out through the top of your head into the sky, and the point it aims at on the celestial sphere is your zenith. When you point straight up, your zenith is what you are pointing at.
What the thread is actually consulting is gravity. The weight settles along the local pull — the direction the planet’s whole mass, summed and averaged, tugs from where you stand. Move, and the tug changes with you. In Denver it points one way; in Perth, nearly the opposite way; at every point on Earth, somewhere slightly different. And the planet is not even a perfect sphere — it bulges at the equator, it is lumpy with mountain roots and dense rock, so the local vertical wanders in ways surveyors must genuinely care about. Two observers in the same city, measuring finely enough, do not quite share an up.
Up is not a property of the sky. It is a relationship between you and a planet — a private conversation with a mass, conducted through the soles of your feet, different at every address.
Notice what has happened to the word. It has not been debunked — your plumb line works; your shelves are level; things still fall down. What has been demolished is only the assumption that your up is the up: an absolute coordinate stamped on the universe. It never was. It was a local reading of an underlying field, and it comes stamped with the address at which the reading was taken. Keep that phrase; the whole paper lives in it.
Now feel the practical weight of the problem, because for centuries it was a genuinely practical problem. Suppose you must map a country. You send out surveyors. Each one plants a tripod, levels it against the local vertical — the plumb line again — sights distant peaks, measures angles with exquisite care. Each surveyor produces a small, precise, local map. And every one of those maps is built on a different up, a different horizon, a different private grid.
Stitching those local sheets into one national map is not clerical work — it is the confrontation, in full, with the problem of this series. Every measurement in every notebook is true relative to the frame it was taken in, and the frames disagree with one another by exactly the amount the planet curves between them. Add the sheets naively — treat every surveyor’s up as the same up — and the errors compound until the map is fiction. The nineteenth century’s great triangulation surveys spent decades on precisely this: not on measuring, which they did superbly, but on reconciling — on the mathematics of carrying a measurement out of the frame that took it.
Two butterflies, two skies, no shared signposts. Two surveyors, two verticals, no shared grid. It is the same predicament — except that the surveyors solved theirs. The solution runs every satellite over your head tonight, lands airliners in fog, and puts a blue dot on your phone that both Denver and Perth agree about. It is one of the most successful ideas our species has ever had, and almost nobody outside the field knows its name.
The solution was not to crown one observer’s up as the true one. Greenwich got the prime meridian, but no city on Earth got the vertical; there is no imperial up. The solution was stranger and more honest: give up on absolute directions entirely, and instead build reference frames — explicit, named coordinate systems, each with a declared origin and declared axes — and become fluent at moving between them. Geodesy settled on three, and the three are worth meeting, because we will meet their semantic twins before this series ends.
The star frame. Origin at the Earth’s center; axes aimed at fixed points on the distant celestial sphere. It does not rotate with the planet — it holds still against the stars, which is why the physics of orbits looks simplest written in it. (The trade calls it Earth-Centered Inertial.)
The planet frame. Same origin, the Earth’s center — but its axes are bolted to the planet itself: one along the spin axis, one through the equator where it crosses the prime meridian. It turns as the Earth turns, so that every city, mountain, and buried cable keeps a fixed address in it. This is the frame your GPS coordinates live in. (Earth-Centered, Earth-Fixed.)
The personal frame. Origin: you. Axes: your east, your north, and — there it is at last, given its rightful, humble place — your up, the local plumb line, one axis among three, valid precisely where you stand. (The topocentric horizon frame.) Denver has one. Perth has another. Every observer who ever pointed at the sky was pointing inside their own.
Here is the discipline that makes the whole scheme work: a measurement is never just a number; it is a number plus the name of the frame it was taken in — and geodesy provides the explicit machinery for carrying a measurement from any frame to any other. Where you stand — your latitude, your longitude — is exactly the information needed to convert between your personal frame and the shared planetary one. The conversion is bookkeeping: a shift of origin, a turn of axes. The insight is not the arithmetic. The insight is that the arithmetic must be done — that a raw number carried naked across frames is not data but noise.
Watch the machinery run once, end to end. Denver sights a satellite: so many degrees above her horizon, so many degrees east of her north. Perth, same instant, same satellite: entirely different numbers, above a horizon tilted 166 degrees from hers. Raw, the two reports flatly contradict each other. It is the 4.4-percent regime of the previous paper — coordinates swapped between frames that share nothing, arriving as nonsense.
Now do it properly. Denver takes her reading and, before comparing it with anyone, re-expresses it: knowing her own position on the planet, she rotates and shifts her personal-frame numbers into the shared planet frame. Perth does the same with his. And in that common frame the two reports become one point — the same satellite, at the same place, to the precision of their instruments. The contradiction is gone, and notice what was never required: Denver did not adopt Perth’s up. Perth did not adopt Denver’s. Nobody’s vertical won. Each kept a private frame and learned to declare it.
Agreement was never about sharing a viewpoint. It is about sharing a frame of reference and honestly declaring where you stand within it. The surveyors did not conquer subjectivity. They routed around it.
This is, I think, the most transferable single lesson our species has extracted from a physical problem — and the reason I have made you spend a paper on rocks and strings. But one piece of the miracle is still unexplained, and it is the piece everything else rests on. The scheme requires a shared frame. Shared frames need a shared origin. Who chose it? Who could choose it, on a planet where every observer’s instruments are local?
Return to the plumb line — the instrument that consults nothing visible. It holds one more secret, the best one. Denver’s thread points one way. Perth’s points nearly the opposite way. Cairo’s, Quito’s, Reykjavik’s: every plumb line on Earth aims somewhere different. Now extend every one of those threads downward, through the ground, and follow them.
They meet.
All of them — every private, local, contradictory vertical on the planet — converge, to a very good approximation, on a single interior point: the Earth’s center of mass. The centroid, in the physicist’s word: the average position of every kilogram of the planet, the point around which all of it balances. That is the point the great reference frames plant their origin at, and now you can see why the choice is not a convention at all. It is the one point every observer’s own local measurement already secretly agrees about. Your up and Perth’s up contradict each other absolutely — but your down and Perth’s down, followed far enough, arrive at the same address. The disagreement was only ever at the surface. Underneath, every private frame was aimed at a common heart.
The shared origin was not legislated. It was discovered — sitting at the average of everything, reachable from inside every private frame, by every observer, using nothing but what they already carry.
No committee could have imposed it; no committee was needed. The center of mass is findable by physics alone from anywhere on the surface — which is precisely what qualifies it to anchor the frames everyone shares. Hold that thought hard. Hold it for three more papers. When this series finally builds a shared frame for meaning, the origin will be found the same way — not chosen by anyone, but computed as the average of everything, the balance point of the whole cloud. The mathematics will even keep the name: the centroid. But I am ahead of the walk.
Now carry all of it back into the box.
An embedding — a word’s star, hung in the universe inside the box — is a local reading of an underlying field, stamped with the address at which it was taken. That is not an analogy I am decorating the mathematics with; it is what the object is. The coordinates of stone in one model are its position relative to that model’s origin, along that model’s axes — a private vertical, valid inside one universe, meaningless carried naked out of it. Every model is a personal frame. Every star chart from the previous paper was drawn in someone’s topocentric coordinates. And the unlabeled axes I kept pointing at — the sky with no signposts, no north, no printed grid — now have their proper name: a universe in which nobody has yet built the shared frame, where every observer still points at their own zenith and calls it up.
So the question that closed the last paper — how could two butterflies ever determine they are describing the same structure? — is the surveyors’ question, wearing different clothes, and geodesy’s answer sketches the shape any solution must have. Not: crown one model’s coordinates as true. Not: force every mind to think in the same axes. But: find the frame-independent things — the quantities that, like the angle between two mountains, come out the same from every valid viewpoint — declare where you stand, and carry every measurement explicitly across. Agreement without surrender of viewpoint. It has been done once, at planetary scale, by people with rocks on strings.
But an honest walker points at the gap in the path as well as the path. The surveyors had something the boxes do not obviously have: a planet. One shared physical mass, tugging on every plumb line, donating its center as a free origin and its spin axis as a free direction. Two embedding models trained in different buildings, on different text, by different organizations share no silicon, no training run, no committee’s convention. If a shared semantic frame is to exist, something must play the Earth — some common mass that both universes formed around without ever touching each other. Whether such a thing exists is a real question, and I will not pretend the arm-waving settles it. I will only note what both boxes were pointed at while their skies condensed: the same world, described by the same species, in the ten thousand shadows of the same human language. Whether that is mass enough to hang a frame on — whether meaning has a center that every private universe’s plumb lines secretly aim at — is, from here on, the empirical question of this series.
Before we can build the shared frame, though, there is wreckage to inspect. Because here is the uncomfortable arithmetic of everything this paper established: if a measurement is only meaningful with its frame declared, then what is the status of every measurement taken before anyone checked the frames? Inside the box, similarity scores are being read off by the trillion, every hour, between vectors whose frames nobody aligned — angles measured against two different ups, added together as if they shared a sky. The next paper walks into that wreckage: what the measurements have actually been measuring, why the model itself never noticed, and how far an origin can quietly drift before every ruler in the universe is lying at once. The walk continues.
Refutations welcome