Hold The Horde — Paired Transfer Discovery Detail

Status. One M4 detail sample: Two Language Discovery Map's chain 3, the visitors' paired transfer register, worked to the depth a writer and a level author could build from. It adds no room, mechanism, encounter, person or objective, answers no central mystery, and fixes no story as canon. Everything new is a string or an occurrence inside an existing host, and each is flagged where it appears. All five owner questions remain unanswered.

1. The discovery, as it plays

The first bay is a machine doing something obvious. A screw head advances into a line in slow strokes; forty paces away across the same room a vessel fills, its level climbing in steps that match the strokes. Nothing is labelled in any human hand. Beside the screw head, cut into the plate, is a bracket mark and a short token; beside the filling vessel, the matching bracket and a different token.

That is the whole of it, and it is enough to start: two marks that belong together, and a machine that tells you which end is which. The player did not read anything. They watched a thing move and noted which mark stood at the moving end.

The second bay repeats it. Same two tokens, same roles, different bracket. Now there is a pattern worth a note, and the natural note is the wrong one: the half-mark that ends in the first form is the source. Both bays support it. It is tidy, it is available, and it is a coincidence.

The third bay is where the design earns its keep. The machine drives from the end carrying the second half-mark. A player who wrote down the suffix rule has a flat contradiction and no way to explain it. A player who wrote down the token at the driving end has three for three. The correction costs nothing but the walk — and it removes a method rather than reversing an answer, which is the harder and more interesting kind of being wrong.

Then, much later and somewhere else entirely, the player finds one half of a fourth pair beside an untraversed opening, and the other half a long way off beside a fitting they have already seen. They can predict that these two are one line. They cannot predict that a person can walk it.

2. Occurrence and evidence inventory

All hosts exist in Station Room and Module Inventory; all identities are reused. A-TRANSFER is the crosswalk's index alias for the shared observation.

Newly proposed strings — five, all inside existing hosts: the role tokens NEB and KUR, and the bracket forms MEV, ANK, TEL. The supplied tokens OSH·a, OSH·b, SORAN, KETH, VEL, TAAN are preserved unchanged and none is reassigned. Nothing here is presented as pre-existing that is not.

3. Inscriptions and physical tells

Transliteration only — no glyph art, image, model or voice is proposed, and the cut forms themselves are for a later art pass.

The marks, as the player would transcribe them.

The bracket form (MEV, ANK, TEL, OSH) says these two belong together. The half-mark (·a, ·b) distinguishes the two halves and carries no reliable direction. The role token (NEB, KUR) is the only mark that tracks the observed driving and receiving ends, and it does so at every one of the four pairs.

Physical tells, so the direction is watched rather than read. At a driving end: a screw or ram head advancing in visible strokes; a check flap that opens only outward; a pressure tell rising downstream of it and never upstream. At a receiving end: a level climbing in steps that match those strokes; a vent puffing once per step; an overflow lip with dried residue below it, evidence that this end has filled before. At R18 the existing exposed non-reversing ratchet says the same thing about that fitting in a second, independent way.

Record descriptions the journal may attach, each a factual observation with its place and life, none an interpretation: "screw advancing, plate reads MEV·a NEB"; "vessel filling in steps, collar reads MEV·b KUR"; "sorter feed driving, plate reads TEL·b NEB". The journal never writes source, destination or direction.

4. Interpretations, tests and outcomes

The two rivals that actually compete after two pairs.

• Suffix reading: the ·a half is the driving end. Supported by R21 and R22, contradicted by R23. It is available, tidy and wrong, and nothing in the interface discourages it.
• Role-token reading: NEB stands at the driving end, KUR at the receiving end. Supported by all four pairs.

Both readings agree on which two marks are one pair, so both send the player down the same continuation at R15. The route test cannot adjudicate between them — only operating a pair and watching can. That separation is the point of the chain, and the interface must not blur it.

Other rivals the source names stay live and are not refuted here: pair identity without direction; ownership rather than flow; the pair marks material flow but not passage; visually similar brackets being unrelated.

The third pair. If the player generalised from the suffix, R23 is a correction: their rule fails against a machine they can watch as long as they like. If they used the role token, R23 is redundancy and confirms it. Either way nothing is graded, no badge appears, and the earlier note keeps its original wording in revision history.

Outcomes of the route attempt at the R15 continuation.

The pair observation at R21/R22/R23 is unaffected by every one of these. A complete local transfer observation remains valid when a route application fails — the player learned a role rule from machines and a walkability fact from walking, and only the second one was tested.

5. Orders, missing evidence, help and later use

Language-and-network first (1→5→7→3→6→9→4→8→11→2→10→12). Objective 5 comes before 7, so the player meets OSH·a at R15 beside a crawl before any pair teaches them what a bracketed half-mark is. It reads as one more unexplained mark. Objective 7 then supplies three pairs and the role rule, and the R15 record becomes retroactively legible — the reinterpretation the design keeps asking for. Objective 6 follows, and R18's OSH·b completes the pair.

Build-and-ship first (1→2→3→4→8→6→10, then 5→7→9→11→12). The order is inverted: objective 6 comes early, so R18's OSH·b is seen before the pairs exist, and R15's OSH·a is not seen until objective 5, late. The player accumulates two halves of a pair they cannot yet recognise as a pair, then meets the pairs afterwards. The chain still completes, and the route attempt is available whenever both halves are recorded; what changes is how much of the work is recognition rather than discovery.

Evidence out of order or missing. With one pair the directional hypothesis is unsupported and the interface says nothing; the player may still attempt the continuation on a guess and get a real arrival record. With two agreeing pairs the suffix rule and the role rule are indistinguishable, which is the intended trap. With R23 only and neither earlier pair, there is one observation and no rule. If the R15 half is never found, the far-bay OSH·b stays an isolated mark; if the R18 half is never found, the R15 continuation is a route with no predicted destination — the player can walk it as ordinary exploration. No objective is gated on any of this: Regional Discovery Map keeps the authored long route and gauges viable, and objective 6 completes from gauges and long access without language.

Optional survivor help, requiring no unselected first survivor. Whoever is aboard responds only to a record the player deliberately shows them, and only from their own basis. A survivor may say that crews used the maintenance crawls, or that a bay of that kind received rather than sent — attributed testimony with its stated basis. No survivor reads either script, none knows current topology, and every chain-3 step is completable with nobody aboard. Nothing is withheld to protect the deduction.

Later use. The role rule makes objective 6's redirected pressure line and objective 11's inward receiver more predictable. At R36 the existing OSH·b sits beside an existing inward flow arrow, and a player holding the role rule has a consistent expectation there while a suffix-only reader has no rule left after R23. What that third OSH·b occurrence means is deliberately not answered here — repeated pair identity, a class mark, or a different fitting reusing the form are all open, and settling it belongs to the language map, not to this sample.

Time and provenance. Framing and committing a capture, operating a fixture and traversing a continuation are world actions with oxygen and threats running. Opening the journal freezes enemies, oxygen, projectiles, readiness, mechanisms, warnings and work progress and completes none of them; comparison and annotation happen there. The display refuses to translate a token, to mark a note right or wrong, to name a direction the player has not watched, or to show a route not walked.

6. State, cues and recovery

Cue text examples (plain, no verdicts): "Plate: MEV·a NEB. Screw advancing." — "Collar: TEL·a KUR. Vessel filling in steps." — "Arrived elsewhere: terminal outlet at [named anchor]." — "Attempt incomplete: branch unresolved."

Recovery choices. A wrong reading is corrected by walking to another pair and watching it, which costs a trip and no progress. A failed route costs oxygen, exposure and possibly the expedition's remaining objective window; carried property still returns on death, so no equipment loss is added. A shift makes the walked continuation historical and a fresh check necessary; the pair observations and the role rule survive untouched, because they were never claims about geography.

Materially distinct cases this sample covers — twelve: one pair only; two agreeing pairs; three pairs with the suffix rule broken; three pairs with the role rule confirmed; arrival at the linked destination; arrival elsewhere; incomplete by blockage; incomplete by retreat; incomplete by death; revisit after a shift; the far half missing; the near half missing. What remains later: every string occurrence across the whole station, the other three chains at this depth, the full cross-register relationship with chain 4, exact fixture geometry and timing, the R36 OSH·b question, and any art or audio.

7. Content and authoring requirements

Exactly what is new, deduplicated against first-pass material: five strings (NEB, KUR, MEV, ANK, TEL); eight plate occurrences across five existing hosts (R21×2, R22×2, R23×2, R15×1, R18×1); one authored asymmetry (R23's suffix flip); five physical tell types (advancing head, check flap, stepped level, vent puff, overflow residue) reused across all four pairs rather than authored per site; and four record-description templates. No new room, module, mechanism, encounter, creature, person, scene, objective or portable object. Every host, the R24 framing position, the R15 crawl, the R18 ratchet and the R36 arrow are existing first-pass content.

Added burden that is real: the R23 asymmetry must be authored deliberately and must survive later editing, or the correction silently disappears; the role tokens must be legible at the distance a player actually stands; and the eight occurrences must each be capturable under the existing encounter placement at R23 without the Veilbearer's sheet making a station uncapturable.

A cheaper authored alternative. Drop the role tokens and let the half-mark carry direction consistently — ·a always drives — with the third pair as pure redundancy. That removes three strings, the asymmetry and the correction, and makes the chain a two-observation confirmation. Its discovery cost is the whole middle of the chain: there is no wrong rule to hold, no moment where a tidy note fails against a machine, and no reason to visit the third pair except completeness. What survives is the genuinely good part — direction learned by watching rather than reading — and what is lost is the only place where the player is wrong in an interesting way. The proposed version retains the correction step at the cost of additional authoring; the simpler alternative remains available.

8. Coverage and limits

This is one chain of four at sample depth, not a language specification. It establishes no meaning for OSH or any other token, does not say why the visitors marked anything, does not resolve the third OSH·b occurrence, and fixes no story as canon. Nothing here shows that the trap is fair, that the correction lands, that the tokens are legible or that any of it is enjoyable; those need later owner-authorised implementation and observation with real players, and none has occurred.