**Status.** A provisional proposal for review, not owner-approved. It works one expedition in which two alien languages both matter, showing what the player observes, believes, tests and decides. Token strings are readable text placeholders; no glyphs, alphabets, art or vocabulary catalogue are proposed. Who the builders and visitors were is replaceable candidate fiction, and the episode depends on no central-mystery direction.

**Builders' runs.** Left by whoever built the structure, above and beside fittings and doorways, as short ordered sequences: `KETH SORAN VEL`. The structural rule available to the player is positional — the first token is what the run is about, and the rest qualify it. The same tokens in another order make a different statement. A run is read where it stands.

**Visitors' halves.** Left by a later species, in a visibly different hand, on the builders' surfaces. A statement is split across two places, each half carrying a matching bracket token — written here as `OSH·a` and `OSH·b`. A half alone is not a statement, and a completed pair asserts a relation between its two locations. This language can only be read by travelling.

The difference is not what they are about. It is that one is read in place and the other cannot be.

A junction with three ways off it.

- Above the left way: `KETH SORAN VEL`

- Above the middle way: `SORAN KETH VEL`

- Above the right way: `TAAN`

Beside the middle doorway, low down and in the other hand, a mark ending `OSH·a`. Nothing near it matches.

Capturing this is world-time work; reading back what is captured is paused, as the owner requires. Records keep location and visible context, which here is not decoration — the surroundings are half the evidence.

- The runs are place names, and the tokens are syllables.

- The runs describe what each way is for.

- The order is stylistic and means nothing; two runs with the same tokens say the same thing.

- The single-token run is unfinished, or damaged.

- The `OSH·a` mark is graffiti, or the same language written badly by someone in a hurry.

None of these is silly, and the article does not rank them for the player.

The two shared-token runs are the lever. If runs were place names, two different destinations sharing two of three tokens in swapped order is strange. If order carries meaning, they are a matched pair of statements about related things.

That is a reason to look, not a conclusion. The **test is voluntary and physical**: walk a way and see what is there. It costs oxygen and exposure like any traversal, and it can go wrong.

Walking both: the left way ends in a pump hall with cable trunking running through it; the middle way ends in a cable junction with a pump feeding it. Each holds both things, and what differs is which one the place is *for*.

That fits the positional rule and nothing else on offer — `KETH` first over a way serving pumps, `SORAN` first over a way serving cabling, each run's second token naming what else is present.

The right way's single token, `TAAN`, over a stair the player has already used and knows leads to a vestibule with nothing in it but the stair.

Under the place-name reading, a one-token name among three-token names is odd. Under the positional rule it is exactly right: a bare subject with nothing to qualify — a place that is *about* only itself.

This does not prove the rule. It is a third instance consistent with it and awkward for the rivals, which is the honest state of the evidence, and enough to make a decision on.

Much later, deep on the far side of the station and reached the long way round, the player finds `OSH·b` beside a bay whose own builders' run begins `SORAN`.

Now both halves exist and the pair can be read as a statement: these two places stand in some relation. It does not say what relation, and it does not say how to get from one to the other.

But the builders' run at that bay says it is a place *about* cabling, and the junction's middle way says the same. Two independent languages, one saying what a place is for and the other saying these two places are related, together support a claim neither makes alone: the middle way probably reaches this bay.

**Neither language is redundant.** Without the runs, the pair relates two anonymous places. Without the pair, two places are merely about the same subject, which is true of many places.

The player can take the middle way, which they now predict comes out at the far bay, instead of the long route they have walked before. Shorter, less exposure, less oxygen.

If the reading is wrong they find out by arriving somewhere else, and they see where they actually are — information rather than a refusal. They have spent the attempt's oxygen and must fall back on the known route, subject to the station's state and their supply.

A wrong *rule* fails the same way and teaches more: a player who decided the last token is the subject predicts differently, walks, and finds the prediction broken in a specific place with the run still on the wall above them.

Runs are plural and persistent, so none is load-bearing. Any two sharing tokens in different order support the same comparison, and any short run among long ones does the work `TAAN` does. Visitors' pairs recur, and a player who completes none still has the runs, which support a weaker but real version of the route decision.

The design does not claim universal access. A player who never compares occurrences may not reach the positional rule, and the survivor route below does not fix that, because it hands over referents rather than the rule.

Yeva worked this spine for years and can say what the crew called two of the ways: the pump road and the cable road. That is true human practice, honestly reported, and she says plainly it is what the crew called them and not what the writing says.

It gives the player referents for `KETH` and `SORAN` at a stroke.

**What remains after asking.** The positional rule is untouched: knowing one way is the pump road does not say whether the first or last token names it, and the player must still compare the runs to see order doing work. `TAAN` stays unexplained. The visitors' hand stays wholly unexplained, because Yeva never read it and says so. The route inference still has to be made, and can still be wrong.

Help shortens the road without walking it, and the unaided player is not treated as the only legitimate one.

Proposed: two languages distinguished by structure rather than subject matter; a positional rule and a paired-halves rule small enough to infer from a handful of instances; physical traversal as the test; and a route decision that both languages contribute to.

Aligned with the interface contract: a committed world test may record that a prediction matched or did not match *here*, without ever declaring a token correct. That is the right shape for this episode — walking the middle way confirms what the player found at the end of it, not that the first token is the subject. The rule stays theirs to hold or abandon.

Open, and belonging to [Language and Journal](/Hold%20The%20Flood/Language%20and%20Journal): how records are compared and revised, and whether a completed pair is flagged as related or left for the player to notice. The episode works either way, though the second changes how much thinking remains.

Also open: what these tokens mean beyond the referents Yeva supplies, who the visitors were, and why they wrote across the builders' walls. Nothing here establishes that inferring a positional rule is enjoyable, or that players will test rather than wait to be told.