Hold The Horde — Language and Journal

This page is the journal contract for Language Expedition, extending Clues and Help and Oxygen and Observation. The station shifts on death (Owner Brief), so old inscriptions, locations, route links and arrival records describe the configuration observed at the time and cannot promise current connectivity; a different destination after a shift does not by itself refute a language interpretation.

Record the test, not correctness

The journal preserves what was observed, supports comparison and records a voluntary test. It never grades freeform notes, labels a hypothesis supported/disproved or declares a word solved. For a route link it may record only arrived at linked destination, attempt incomplete or arrived elsewhere, with the actual place and interruption attached. Survivor help is attributed testimony, not system truth.

This reversible feedback separates a completed traversal from an interrupted one while leaving later evidence free to change the interpretation.

The information cycle

Capture, positioning, waiting and manipulation happen while the game runs, with its oxygen/exposure. Opening, comparing and annotating pause the game, including enemies and oxygen during combat. A paused view cannot reveal a new angle, sample a later state, operate a control or complete capture.

Additions to the current journal

The existing framed still, visible context and location remain. This page adds four tools:

1. Comparison board. The player pins records and highlights groups; links say “player-linked.” On-request visual suggestions are fallible recorder assistance, not automatic pairing or meaning.
2. Revision-preserving notes. Editing keeps earlier wording, so contradiction shows learning rather than rewriting history. Freeform text is never parsed for meaning.
3. Prediction/result pair. The player links a route/control to a stored destination/outcome; afterwards the journal attaches the place, state or interruption. Selected records drive it, and it awards no correctness or rule.
4. Attributed help. A response records who spoke, what they did or saw and what is interpretation. One record cannot reveal unseen inscriptions.

Automatic semantic labels, a confirmed dictionary, batch completion and undiscovered-site hints are excluded. All four additions survive return and death as knowledge. A revisit never overwrites an old still or outcome: the player captures a new state beside it, so an earlier arrival remains historical rather than promising the route is unchanged.

Applying the contract to both languages

Builders’ ordered language

At the three-way junction the player captures KETH SORAN VEL above the left way, SORAN KETH VEL above the middle and TAAN above the right. They pin the first two, highlight the swapped groups and may note “first token says what a place is for; later tokens qualify it.” The note is ungraded.

Walking the ways is voluntary and costs ordinary oxygen/exposure. Two visible mains run along this spine. In the left pump hall, the pump main terminates at the pumps while the cable main crosses between walls. The middle way reaches a cable junction and continues past it; there the cable main terminates in banked connections while the pump main crosses and carries on. A small cooling pump also terminates in the junction, but it is visibly fed by a local branch rather than forming the end of the pump main. Token order therefore correlates with which of the two mains ends in the room, not with everything that terminates there. This supports a purpose-and-order reading over meaningless order, while compound place names remain a live rival. The bare TAAN above a stair-only vestibule is another consistent case, not proof. The journal records the places reached; it never marks KETH as “pump,” asserts a grammatical subject or upgrades the note to fact.

Yeva can truthfully call the first two the crew’s pump road and cable road because she worked this spine. Her attributed gloss supplies referents, not the positional rule. The player must still compare order, interpret TAAN and decide whether the pattern applies elsewhere. Multiple persistent runs give an unaided route without guaranteeing every player notices it.

Visitors’ paired language

The player separately records a visitor threshold half ending OSH·a plus a further token beside the middle doorway. Before making any prediction about the far bay, they have recorded two or three other pairs at the ends of transfers whose direction is independently visible: a driving device pushes into a line at one end, while a vessel receives and fills at the other. Across those pairs, one further token recurs at the driving end and another at the receiving end. This small comparison supports a partial local role interpretation — a source/destination distinction in this distributed register — without settling either token’s full meaning or a rule for the entire visitors’ language.

Much later, beside a far bay whose builder run begins SORAN, the player finds OSH·b carrying the receiving-end token, while the junction half carries the driving-end token. They can pin both halves and manually link them. An on-request recorder suggestion may surface the visually similar bracket token, but does not assert a semantic pair, relation or direction; recognising how the further tokens correspond to the observed transfers remains interpretation.

The visitor pair suggests a transfer connection from the junction toward the bay; the builder runs independently suggest both concern cabling. A transfer connection does not prove a walkable route. Beside the source mark, a conduit enters a human-sized maintenance passage continuing beyond the junction; at the far bay, the receiving fitting stands beside another maintenance opening. The player previously turned back at the junction on supply. Those visible accesses make the untraversed continuation worth investigating for a possible connector, without establishing continuity, destination or a shorter journey. They may link that continuation to the far-bay record and attempt it. The test completes only by continuously traversing the continuation and emerging at an identified outlet. Reaching the far-bay anchor records arrived at linked destination. Reaching a terminal outlet at another persistent anchor records arrived elsewhere. Neither event translates OSH, proves every pair is a route or confirms the builder grammar.

Yeva says plainly that she never read the visitors’ hand. Her road names constrain a token-to-place comparison but do not translate tokens; the visitor relation and combined route inference remain the player’s. If OSH·a was missed, another persistent pair can expose the distributed use, while the builder runs alone support a weaker route hypothesis. A completed route spends time and oxygen while the game runs and reveals a destination; a detour may interrupt the work planned for that expedition. Every item still carried comes home on death, so the player may press on, retreat where practical or end the trip by dying and return aboard with that load. A blockage, intermediate junction, unanticipated branch, retreat, interruption, abandonment or death before the defined terminal records attempt incomplete, not “arrived elsewhere”; only a fully traversed candidate with no unresolved continuation yields an incompatible destination. Supply and threats still shape what happens before the trip ends, without requiring equipment loss or a perilous walk home.

The same distinction applies to controls. A powered, intact control with the relevant component can record the visible state after activation. Missing power, a removed component, interruption or an unfinished action records absent test conditions or an incomplete attempt—not evidence against the prediction and not a general diagnostic guarantee.

Withheld correctness versus limited confirmation

Withhold structured result records. The player relies on the world and their notes, with no arrival/test status. This keeps interface authority low and allows ambiguity, but does not uniquely create ownership: explicit help or an observed result can still leave comparison, generalisation and action to the player. Its cost is a possible diagnostic burden; oxygen, damage, missing power and contextual exceptions can make an unfinished test harder to distinguish from a failed reading. Visible environmental evidence remains available and may explain the outcome even without a journal status.

Limited observation and help — preferred. Record the reached place, visible control state or incomplete condition, and keep testimony attributed. On-request visual suggestions among already captured records are kept. Their cost is some reduction in unaided noticing. The reason to keep them is visible in the complete episode: surfacing a lookalike only brings two records together; the player still appraises whether the halves pair, infers the source/receiver distinction from independent transfers, applies it to a new pair and chooses whether to act. Manual comparison remains fully usable. The player still words the hypothesis, judges whether an observation bears on it, reconciles human knowledge and pays the in-game cost of acting. The other costs are outcome states, comparison consistency and a risk that even factual status labels feel more authoritative than intended.

Language Design Evidence describes shipped uses of uncertainty, candidate meanings, repeated contexts and validation. Its multiple-context precedent supports planned redundancy, not a mandatory instance count.

Open questions

• How object identities and relational clues survive the station’s rearrangement on death remains to be designed.
• Reopen the choice to record the test rather than correctness if observation of players shows them treating arrival records as language answers, unable to diagnose attempts, or using comparison as clerical transcription; stronger confirmation or fully withheld structured feedback would then be the alternatives.
• These rules are untested; numbers are proposals.