The problem#

At every moment in a scene, an attentive reader should be able to point. Point at the falls, at the city, at the sky, at the direction the light is coming from. When they can, the habitat feels like a place. When they cannot — and worse, when they believe they can and are wrong — the world quietly stops being solid, and no amount of good sentence-level writing puts it back.

Gyrealm makes this harder than a terrestrial setting does, and the difficulty is structural rather than incidental. A rotating cylinder has three independent axes and no horizon to anchor them. English’s directional vocabulary was built on a planet and carries planetary assumptions inside words that look neutral. The result is that spatial errors here are not slips of attention; they are the default outcome of writing carefully in inherited language. Getting them right takes deliberate method, which is what the rest of this bundle is about.

Two things are at stake and they are the same thing:

• Physical correctness. Does the geometry hold up? Which way does falling water deflect, where is the sky, what is actually overhead.
• Reader mappability. Can someone holding the book build a stable map and keep it across chapters?

They collapse into one problem because in a cylinder the reader has no intuition to fall back on. On Earth, a writer can get a compass direction wrong and most readers will never notice, because the map they are building is anchored by a lifetime of standing on a planet. Here, the prose is the only source of the reader’s map. An inconsistency is not a small error against a robust background; it is a corruption of the sole record.

Observed instances#

The following are all live, in pages currently in the repository. They were found by tokenizing every directional word in the affected pages and resolving each against the habitat frame — the procedure described in The Spatial Ledger.

1. Canon describes two different habitats#

Coriopolis says the cylinder “is split into two distinct sections, separated by a 2-kilometer-wide gap,” and then locates the emitters on “the outer surface of the inner cylinder and the inner surface of the outer cylinder.” Those two surfaces face each other across a radial gap: this describes a nested pair, a smaller cylinder inside a larger one, with the 2 km measured outward from the axis.

Dawn at Scimitar Falls reads the same sentence as an axial split: “two cylinder sections separated by a 2-kilometer gap. Holographic emitters on the facing surfaces of the two sections” — two short cylinders end to end, the gap measured along the length, emitters on the facing end-caps.

These are not two descriptions of one habitat. They are two habitats. There is a single question that separates them cleanly:

Standing in the preserve and looking straight up, do you see the far side of the inhabited surface, or do you see a sky?

Under the axial reading you see land — the opposite side of the cylinder, 20 km up, curving away in both directions, with the sky visible as a bright band 40 km north along the axis. Under the nested reading you see an emitter sky 2 km up and never see the far side of the habitat at all, because there is a second cylinder in the way. The answer cannot be “both,” and every description of the Coriopolis sky depends on it.

2. “Overhead” means three incompatible things on one page#

Landscapes of Coriopolis uses the word three ways in twenty lines:

• “the landscape curves upward in every direction, meeting overhead at the opposite side” — overhead is land. The axial reading.
• “Overhead, holographic emitters on the inner face of the gap … project a simulated sky” — overhead is sky. The nested reading.
• “from the preserve, the city’s lights are overhead” — wrong under both readings. The city is at the northern end and the preserve at the southern end, 80 km apart along the axis. What is overhead of the preserve is the far side of the preserve. The city is up-axis, and would appear as a distant convergence to the north, not as a band across the sky.

The third is the most instructive failure, because it is not a physics error at all. It comes from conflating two axes that Earth lets you conflate safely: on a planet, “far away” and “toward the horizon” are the same direction. In a cylinder they are two different axes, and the sentence picked the wrong one.

3. Dawn sweeps in both directions across six renderings#

The scene prompt specifies that “light sweeps along the spin direction.” West is the direction of rotation, so this means the lit line travels east to west, and dawn arrives at an observer from the east — the Earth-familiar pattern, a sun that comes up in the east. Six renderings, one prompt:

This instance was recorded when east ran with the spin. The convention revision of 2026-08-24 inverted the compass reading of the prompt and, with it, every verdict in the table: none of the six is now off-spec. Tolkien’s internal contradiction — it opened with a ridge sheltering the falls from the light and then sent the light the other way — resolves in the same stroke, since dawn now arrives from the east and the sheltering ridge is the east ridge.

That is not a vindication of the prose. Not one word of the renderings changed its mind; the frame moved underneath them, and four sentences went from wrong to right without being touched. A defect that a convention change can silently repair could as easily have been silently created. The deeper trouble also survives untouched: the prompt picked a sweep direction without recording why. See the open questions below.

4. A number migrated between axes#

Five of the six renderings attach eighty kilometers to the sky: “eighty kilometers of sky that was not sky,” “Eighty kilometers of engineered air,” “eighty kilometres overhead,” “eighty kilometers of engineered atmosphere,” “eighty kilometers of overhead.”

Eighty kilometers is the cylinder’s axial length. Under the nested reading it is a defensible figure for the sky, which would run the full length. Under the axial reading it is meaningless for a sky located at the gap. And in the two renderings that make it a span the light travels across — Tolkien’s and Martine’s — it is simply wrong on any reading, because that span is circumferential, and the circumference of a 20 km cylinder is 62.8 km. Eighty kilometers is not a distance that exists around the curve.

The prompt supplied “eighty kilometers” for the axial distance to the city. Every rendering wanted a scale word for the sky, and the number was the only one in the room.

5. Directional prepositions hide the source/destination ambiguity#

The Le Guin rendering has “the deflected currents pull at them from the west.” This can be read as wind originating in the west (blowing eastward) or as a pull toward the west. The prompt establishes mist that “drifts laterally … against the east ridge,” which requires net eastward transport — the first reading. The second contradicts it. Note which way that went: the 2026-08-24 revision flipped which of the two readings is correct without altering a syllable of the prose. A sentence whose correctness depends on which live English convention the reader happens to apply is not repaired by becoming accidentally right.

English is systematically ambiguous here: a “westerly wind” comes from the west, but a “westward drift” goes toward it, and “pull from the west” belongs to neither convention. On Earth this rarely matters, because the reader has weather intuitions to fall back on. Here there are none.

A defect in the frame itself#

The five instances above are errors against the coordinate system. A later pass found one in it, and it took two attempts to fix — which is the more instructive half of the story.

The symptom. Canon defined north by the right-hand rule on the rotation vector, east as the direction of spin, and up as toward the axis. Each is individually reasonable; together they made the local frame left-handed. An inhabitant standing on the floor and facing north had east on their left — the mirror image of Earth.

The derivation is three lines. Put the spin along ẑ, so ω = ω ẑ, and stand a person at radius R along x̂. The floor there moves along ẑ × x̂ = ŷ, so east is ŷ. Up is −x̂, because up is inward. For any standing observer right = forward × up, so facing north the right hand points ẑ × (−x̂) = −ŷ, which is west.

The root cause is an import. The right-hand rule is planetary practice, and on a planet up points away from the rotation axis; inside a cylinder it points toward it, and negating one leg of a triad flips its handedness. The shipped convention was correct for someone standing on the outer hull of the drum, and mirrored for everyone living inside it.

The first fix, and why it was wrong. North was redefined as the left-hand-rule thumb. That does restore east-on-the-right, it is self-consistent, and it cost nothing to apply, since no external reference fixes which end of a habitat is north. It was also the wrong repair. Handedness was one of two things the inversion broke, and flipping north fixed that one while leaving the other inverted and adding a third: the habitat then turned counterclockwise looking north where a planet turns clockwise, dropped objects drifted west where on Earth they drift east, and eastbound travel made a body heavier where on Earth it makes it lighter.

The actual root. East = spinward had never been examined. It was assumed early and everything else was fitted around it, including the first repair. It is the wrong leg of the triad to hold fixed, because it is the one carrying no weight: nobody can feel the spin. Coriopolis turns at 0.30 rpm, below every vestibular threshold — see turning bias. What residents actually perceive is the deflection package, falling water drifting and weight changing with travel, and naming east with the spin inverts every item in it relative to what a reader knows about Earth.

The fix. East now runs against the spin, and north returns to the right-hand-rule thumb. East is still on the right facing north. The habitat turns clockwise looking north, as a planet does; dropped things drift east and eastbound travel is lighter, as on Earth; and a habitat aligned with the celestial pole spins prograde with the Solar System rather than retrograde against it. No geography moved — the city stays north, the preserve stays south. What moved is the deflection table, which inverted.

Corroboration, noticed after the decision rather than used to make it: the unreconciled source notes in notes/ use this convention throughout — right hand to north, fingers curling west, clockwise looking north, a cascade curving east, and winds “against the rotation, from the west.” Canon had logged three of those as errors. They were not errors. The original material was self-consistent, and the wiki was the deviation.

What makes all this worth recording is why six careful renderings never caught the original defect. Not one of them ever needed a character’s left or right. The error was unreachable from the prose, and would have stayed unreachable until the first scene in which somebody turned their head. But the sharper lesson belongs to the first repair, which passed every check that had been written — because the checks were derived from the same unexamined assumption. Auditing the frame is not enough. You have to audit the leg of the frame you decided not to question.

What is holding up#

The audit is not all bad news, and the part that works is worth naming because it shows the standard is reachable.

The falls deflect one way, consistently, everywhere. Water leaving the lip at a smaller radius carries the tangential speed of that radius; as it falls outward the floor beneath it is moving faster, so the floor outruns it and it lands antispinward — which under the current convention is east. The overlook, the ridge, the pooling mist, and Wolfe’s “one could have said the water chose to go east” all agree with each other and with the physics.

These were written as west and have been relabelled; the physical claim underneath never changed. That is the test of a convention change, and this one passes it.

Assuming a 500 m drop, the deflection works out to roughly 110 m of eastward drift over about ten seconds of fall — dramatic, visible, and consistent with Scimitar Falls calling the curve “visible to the naked eye.” The drop height is an assumption; the wiki does not specify it.

Open canon questions#

These are worldbuilding decisions, not craft decisions, and this page does not settle them. They are listed here because the prose cannot be fixed until they are answered.

1. Nested or axial? Is the 2 km gap radial or along the length? Every sky description in the project depends on the answer, and the two current answers are both in canon.
2. Which way does dawn sweep? The analysis page shows this has three defensible engineering answers, not two, and that the Earth-intuitive one may be the physically natural one. The prompt picked an answer without recording why.
3. How tall is Scimitar Falls? Currently unspecified, and the deflection — the thing the landmark is named for — cannot be described precisely without it.