Hold The Horde — Two Language Discovery Map

Status. First-pass SMALL-band proposal: four finite discovery chains across two alien languages. It is not a conlang, glyph design, final history, story truth, implementation or duration claim. The owner requires at least two alien languages and practical clue-led exploration; the exact authors, vocabulary, fiction and placements below remain replaceable. Placeholder text represents recorded shapes, not English pronunciation or final art.

What the two languages are

The authored world provisionally contains two actual languages, not five languages inferred from five uses. The player encounters bounded registers of each and learns operational rules, not a complete dictionary.

Each alien language changes action beyond naming: Language I predicts facility purpose and candidate wall access; Language II predicts transfer direction and a limited reply/deferral protocol. A player can learn these local rules while remaining unable to translate either language generally.

Four-chain summary

Chain 1 — builders' ordered facility runs

Accessible evidence. Objective 5 hosts the existing three-way junction: KETH SORAN VEL above the pump-hall way, SORAN KETH VEL above the cable-junction way and TAAN above a stair-only vestibule. In the pump hall, the pump main ends and the cable main crosses. In the cable junction, the cable main ends and the pump main crosses; a small local cooling pump also terminates there, so the comparison is explicitly between the two continuous mains.

Reasonable rivals. The strings are compound place names; first position marks primary purpose; order is stylistic; TAAN is a damaged run. None is dismissed by interface text.

Test and authored operational rule. The player physically walks to and records both rooms, then compares which main ends. The authored consistency rule is that first position identifies the primary service subject in this facility-run register and later positions qualify it. That rule lets content be written consistently; the journal never turns it into a translation. Either room may be reached first. One room's observation, an interrupted capture, or a mere old route record lacking the required view leaves the comparison incomplete. Valid committed captures of both mains remain linguistic evidence after death and may be compared even when taken in different lives; their room, life and arrangement provenance remains visible.

Correction and uncertainty. Swapped order tracking different visible mains weakens the stylistic reading. TAAN is a third consistent case, not proof. Compound naming remains viable. A later control that behaves contrary to the player's chosen first-token prediction challenges the generalisation only after power, component and action conditions were valid; a missing part, invalid control state or interruption is not semantic failure.

Practical later use and cost. At objective 6, run order helps choose the control likely concerned with the pressure main rather than the purge branch. At objective 11, it predicts the pump-purpose/inward selector. The player still reads gauges and performs world-time work. Acquiring the two captures costs traversal, oxygen and exposure. The linguistic comparison does not expire; only a claim about a room's present state or a new control test needs current observation and access. Historical locations never become current routes.

Chain 2 — builders' wall-access register

Accessible evidence and placement. This is the existing wall-string chain placed without new rooms: objective 1's service intake contains the string beside a pressure door and its matching blank panel; objective 5's cable-junction edge contains a matching string with a panel already off and a continuous crawl visible; objective 8's decontamination approach contains the discriminating matching string beside an open removable slab and an ordinary standing passage, plus a visibly unmarked duct welded closed; objective 6's jammed-door approach reuses the blank panel and shared far-side lamp/crew-name sightline for the player's chosen bypass test.

Reasonable rivals. Door label, hidden duct, general access, warning and maintenance responsibility all fit early evidence. The shared cut style and one repeated token support classifying these strings with Language I, but technical shorthand remains a live rival.

Test and correction. Removing the objective-6 panel is staged world-time cutter work. Released fasteners and a fully moved panel persist; cancelling loses only the unfinished stage and spent time/oxygen. Finding an opening supports “something traversable may be behind this mark.” The objective-8 standing passage weakens “duct” in favor of candidate access. An unmarked welded duct shows the marks are not exhaustive, but does not by itself refute “marked wall means duct.”

Failure/incomplete states. A collapse, pinch-out, weld or seal is visible where the candidate stops. That lowers confidence in current operability without proving the mark never denoted access. A blocked, abandoned or death-interrupted crawl remains incomplete; seeing the same landmark from two openings does not prove a continuous passage. Destroyed or missed evidence is not restored by the journal. The player may use the jammed door, retreat or pursue another eligible objective if current conditions permit.

Practical later use and cost. The rule changes which wall is worth opening and may bypass part of objective 6's long approach. It never guarantees current connectivity. In the build-first order, objectives 1 and 8 can supply the broad access reading before objective 6 even if objective 5 is later. In the language-first order, objective 5 supplies the open crawl before the objective-6 test. Dorn may truthfully describe crew use when shown a record, but no survivor is required and he cannot name a currently open route.

Chain 3 — visitors' paired transfer register

Accessible evidence. Objective 7 contains three already-proposed transfer pairs at existing one-way fixtures. At each, physical operation independently shows a driving device pushing into a line and a vessel receiving/filling at the other end. One further token recurs at driving ends and another at receiving ends; paired halves carry matching bracket forms such as OSH·a and OSH·b. One receiving fitting also supplies the cross-register recurrence described above. Far away, a recorded OSH·b sits beside a receiving fitting and maintenance opening in a bay whose builders' run begins SORAN; the corresponding OSH·a is at the cable junction beside an untraversed continuation.

Reasonable rivals. Pair identity without direction; ownership; source and destination reversed; the pair marks material flow but not passage; visually similar brackets are unrelated. Physical drive/fill direction decides only the local end roles.

Test. After comparing at least two complete pairs, the player selects a prediction linking the junction continuation to the far-bay record and traverses it in world time. Reaching the far-bay anchor records arrived at linked destination. Reaching a different terminal anchor records arrived elsewhere with that place. Blockage, branch not resolved, retreat, abandonment or death before a terminal is attempt incomplete. None is a language correctness badge.

Correction and retained uncertainty. A third visible transfer corrects an accidental reversed reading or remains redundancy if the first two agreed. Arrival elsewhere challenges the route application, not the observed transfer roles: conduits need not be walkable, and a death shift may change unbound adjacency. The player still does not know what OSH says, whether every pair is a transfer or why the visitors marked it.

Practical later use and cost. The directional rule makes objective 6's redirected pressure line and objective 11's inward receiver more predictable. The authored long route and gauges remain viable without it. Testing an uncertain continuation spends oxygen, exposure and possibly the expedition's remaining objective window; carried property still returns on death, so no equipment-loss penalty is added.

Chain 4 — visitors' responsive utterance register

Accessible prior evidence. Objectives 5 and 7 place five pulled/intact A/B contrasts. Their three existing builder-workstation cases put a token from that workstation's fitting in the utterance's first/address slot, supporting an address reading. The one newly proposed objective-7 recurrence instead puts a receiving-fitting token in a body/work slot to connect the paired and utterance registers; it is not one of the three address precedents and does not increase their count. Objective 8's extra plate and the opening berth token, if captured, are optional redundancy/cues. At objective 9, the existing regulator gallery has four regulators and a panel: R1 pulled/A with the pry scar cutting through its words and a once-cut retaining crimp; R2 pulled/A; R3 intact/B; R4 intact/A. R4 is the only recorded A beside untouched work.

Rivals. A means completed, requested/next, or foretold; B means incomplete, refusal or deferral. The scar dates R1's text before its pull and weakens “completed record,” but does not distinguish request from foretelling or queued job.

Discriminating actions. Pulling R4 is staged world-time work. Its physical tell drops, and on returning to the panel the player can capture a new line addressed to the berth, naming R3 and ending A. This proves responsive updating, not a mind. The player then unclips R3's own B plate and seats it in the reply face without touching R3. The valid matching plate produces exactly one acknowledgement line and quiets the current line. An A plate, a B plate for another fitting or an external plate produces nothing and remains reclaimable; that silence is a control result, not a semantic judgement.

Exact persistence/deferral rule. The reply face holds one plate. It acknowledges the correct current fitting's B plate once per fitting, ever. While seated, it defers only that incomplete work. Reclaiming the plate makes the existing line current again without writing another; reseating it quietly restores deferral. Completing the fitting closes its line and advances to the next intact fitting whether or not deferred. Completed pulls, written lines and a seated plate persist; removal never reopens completed work. The console can accept word-shaped input without being alive or corresponding to anyone.

Wrong order, missed evidence and a specified later variant. Pulling R3 before R4 removes the next intact fitting: R4 still yields its part and physical result, but there is no R3 line to answer, so the clean B test is lost at this gallery. That is an evidence cost, not a hidden penalty; no part, basic seal action or ending is withheld. A missing scar leaves chronology less certain, and neither scar nor plate is regenerated.

Objective 11 proposes a small same-chain variant inside its existing iris mechanism. On the reachable approach side, a local inlet B plate is clipped to the inlet fitting and bears the same fitting token named by the panel's current A-line. An adjacent reply face accepts one plate. Seating this local matching B plate before equalisation produces the same once-only acknowledgement and defers that current line; reclaiming it restores the same line without another acknowledgement while the work is incomplete. A gallery B plate or any other fitting's plate is invalid and produces nothing because its fitting token does not match. The existing iris work supplies the completion event: set pump/inward and fill the receiving vessel until the two gauges visibly equalise. That committed equality closes the inlet line whether the plate is clipped, carried or seated; the subsequent hand-cycle opens the iris as a separate existing action. The plate has one location—clipped to inlet, carried, or seated—and the panel retains current/deferred/closed plus acknowledgement-unused/spent. A carried plate returns on death; a seated plate and completed equality persist at the iris. This adds one local plate, one reply-face input and state variants to objective 11, not a room, region, mechanism slot or required path.

Practical later use and cost. At objective 11, the current A-line gives one more reason to inspect the inlet; its local B plate can defer only the line while every physical control remains available. The player may instead equalise, hand-cycle or ignore it. Pulling gallery regulators in the readable order preserves the richest earlier evidence; stripping immediately yields the same ordinary parts/work but can forfeit that evidence. The later local case is a concrete second test, not guaranteed recovery: it too can be missed, stripped or left unreachable in a current arrangement. No unseen penalty, required survivor, basic-ending loss or restored clue is added.

Dependencies across all twelve objectives

Both retained orders. In build-and-ship first (1→2→3→4→8→6→10, then 5→7→9→11→12), physical tells and long routes complete 6 before language; 1/8 can establish the broader wall-access reading, and 5/7 later reinterpret stored evidence. In language-and-network first (1→5→7→3→6→9→4→8→11→2→10→12), Chains 1 and 3 inform the riskier objective-6 continuation before the build rewards. Objective 1's recurrence plus objective 5's open crawl support a plausible duct/crawl reading for that attempt; objective 8 later supplies the non-duct standing-passage discriminator, so it is not redundant for the broader “candidate access” inference even though it is not required to try the crawl. An intrinsic sequence remains only within a chain—for example, a transfer prediction requires recorded ends, and the gallery's richest B test requires R4 before R3. While that evidence is incomplete, the player can pursue any eligible objective or use the physical alternative; all progression is not held behind it.

Two discovery-to-action sequences

Two languages at the pressure-receiver iris

At objective 11, the player records a builder run on two selectors and visitor OSH·b beside an empty receiving vessel with an inward flow arrow. Chain 1 predicts that the pump-leading builder run marks the pressure-purpose selector. Chain 3 predicts feeding the visitor-marked receiver rather than purging it. Chain 4's current inlet line and locally clipped matching B plate may agree but are not needed. The player selects pump/inward in world time; the vessel fills, the gauge rises and visible equality completes the inlet work before the separate hand-cycle becomes available. If they select purge/outward, the gauge falls and a visible vent lights. Return to neutral, reverse the selector and continue; the observation corrects the control choice without a wrong-answer label or erased work. A player with neither language can use the reachable two-clamp mechanical line at greater exposure. The languages combine to justify a confident action, not to become universal keys.

“A means already done” meets the gallery

The player's plausible note says A marks completed work. R1's scar through pre-existing text and its once-cut crimp contradict a simple after-the-fact record. They revise or retain the note themselves. Pulling R4 then produces a new current line naming still-intact R3. That visible before/after further weakens “already done.” Seating an invalid plate produces nothing but does not tell the player why. Seating R3's matching B plate, while R3 remains physically unchanged, produces the once-only acknowledgement and quiet. The strongest remaining rule is operational—A-like lines can name current work and matching B can defer it—not a literal translation, proof of a request or proof of a living speaker.

Information, help and continuity contract

• World time: framing/committing a capture, reaching a new viewing angle, operating a control, seating/reclaiming a plate, opening a panel and traversing a test all run with oxygen and threats. A completed physical checkpoint persists under its owner.
• Paused journal: comparison, zoom, manual highlighting and personal notes fully pause the world. They cannot advance a capture, poll a later state, move a control, seat an object or complete a route.
• Predictions and records: the player deliberately links selected records to a route/control prediction. The journal may attach only factual outcomes—actual terminal reached, visible control state or incomplete condition—with source, room instance and arrangement/life. It never parses notes, assigns correctness, fills a dictionary or unlocks a language.
• Suggestions: on request, fallible visual matching may surface only already captured records and is labelled software-generated. It asserts no relation, register, direction, meaning or unseen site. Manual comparison remains available.
• People: a survivor responds only when shown a record. Statements retain speaker and basis: Yeva may name crew roads; Dorn may describe crawl-mark use; neither reads an alien language or knows current topology. Help is not withheld merely to preserve an intended deduction, and no person is required.
• Shifts and objects: after death, carried property returns, completed actions and placed objects retain their owning states, and old route/adjacency records become historical. A new arrival elsewhere after a shift cannot by itself refute a language rule. Consumed inputs remain spent; missing/destroyed evidence is not silently repaired.

Different reasoning paths are valid: a player may infer first-position purpose from rooms, accept an attributed road name as a referent, follow gauges, or test a wrong selector and correct from its physical result. The interface reports observations and limits, never an oracle's preferred deduction.

Scope ledger, remaining work and smaller alternative

This map contains exactly four chains, two per proposed language. It reuses the objective-5 pump/cable/vestibule comparison; six wall-access evidence conditions; three objective-7 transfer pairs plus one far-bay continuation; five earlier A/B cases, three first/address-slot comparisons, the optional berth-token cue and the objective-9 four-regulator/reply-face scene; and objective 11's iris application. The separate new objective-7 body/work-slot recurrence does not change those prior counts. New first-pass content is the explicit two-language/register relation, objective placement of wall-access cases, two cross-register relationship observations at existing sites, one local objective-11 inlet B plate plus reply-face input and current/deferred/acknowledgement/closed state variants, the twelve-objective dependency matrix and reversed-order handling. It adds no region, objective, room module, mechanism slot, equipment part, survivor or mystery chain.

The future 40-room/12-mechanism ledger must assign each cited state to actual placed instances, count the wall panels, transfer fixtures, gallery and iris controls once, and count the iris's local plate/input/state variant within that existing mechanism. It must ensure current access/recovery in permitted arrangements. The five-people/fifteen-scenes ledger may overlap optional Yeva road-name and Dorn crawl-use responses with their existing home follow-through; materially different record states still need counted dialogue variants. This article does not allocate all rooms, write final dialogue, supply glyph assets, enumerate every string or clue permutation, establish all reversed/missed combinations, finish either language, choose a cast/story/ending, or prove accessibility and comprehension.

Later finite authoring must define the exact sign inventory and occurrence ledger, every current-room evidence placement, capture/control failure variants, provenance text, accessibility equivalents, survivor knowledge states and cross-article interfaces; then independently check that each actual chain has repetition, a discriminating case, a physical alternative and no circular gate. Implementation, tuning and player observation require later owner authorization.

Strongest smaller viable alternative. Keep only Chain 1's ordered runs and Chain 3's transfer pairs as the two language experiences; leave wall strings and utterances as unrelated local mysteries. This still provides two real alien languages and two useful predictions while preserving physical alternatives. What it loses is not claimed hours or playtime: Language I no longer changes wall-opening decisions, Language II never becomes something the player can answer or defer, cross-register reinterpretation disappears, and the proposed SMALL four-chain allocation is unmet. Prefer it if the extra registers read as bookkeeping rather than discoveries.

No paper map establishes fun, comprehension, production feasibility, campaign duration, guaranteed recovery or that no player can become stuck. Those require complete authored placement and later owner-authorised implementation and observation.