nimmerverse/nimmerworld.eachpath.local
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

political-register/imperial-narrative.md v0.1: claim named-but-unwritten territory

Browse Source 
Author the imperial daily-narrative paper that has been referenced as
*forthcoming* from political-register/architecture.md:235,
political-register/economics.md:173, identity-and-personhood/vocations.md:97-99,
style/consent-discipline.md:77-79, and architecture-index.md.

Composes existing canonical primitives without re-introducing them:
- Compositor as narrative-composer (per narrative-composition/architecture.md)
- Three-tier policy loop + dual-ledger demands != actions (per political-register/architecture.md)
- Diegetic relays + ring-of-access gradient (per topology-and-rendering/architecture.md)
- Boundary-zone-with-inner-zone audience-architecture (per runtime-engine/event-zones.md)
- Universal translator + dialect-framework + decryption-minigame substrate
  (per player-experience/architecture.md)
- 24h window-partition + shift-start firing (per identity-and-personhood/vocations.md)
- Trait-palette + hex-canon discipline (per style/trait-palette.md)
- Broadcast-as-specter + temporal-grammar (per style/consent-discipline.md)

Net-new substrate (narrow):
- Two-cipher discipline (eye-strict for written/decoder-trainable;
  ear-loose for spoken/anti-monotony) -- modality-aware redundancy parallel
  to the existing color + motion-signature two-channel discipline
- Speech-pattern as typed catalogue-entry on the action-letter (preserves
  catalogue+tools typed-contract from narrative-composition.md; provenance
  flows end-to-end into linguistic-rendering)
- Bilingual-signage divergence-class typing (sanitization | omission |
  misdirection | erasure | outright-lie | null) as gameplay-content;
  organic Rosetta-stone training corpus via signage-exposure
- AI-author delegation as load-bearing-by-design: the partnership-
  architecture's three-way split (substrate / studio-content / AI-author)
  formalized; daily imperial-narrative generation delegated to Chrysalis-
  now / Nyx-future as the AI-collaborator's domain
- Disciplines for daily-generation: register-fidelity (LLM-corporate-fluent
  not Stalin-authoritarian); divergence-discipline (typed not random);
  cycle-coherence; world-state coupling; two-cipher integrity; slum-tongue
  voice-protection; hex-canon propagation; substrate-content boundary

Philosophical register locked: Kafka-indifferent-totality not Orwell-
malicious-state per political-register/architecture.md:11. Imperial-tongue
is the residue of compliance-optimization, not deliberate adversarial
cipher; the gap between what broadcasts sound like and what they optimize
for is emergent. Decoder-progression's terminus is the diagnostic ("the
machine speaks like an over-aligned corporate AI assistant; that is the
violence") not the catharsis. Memorialists and Aletheia-progression are
preserved as emergent (per substrate-vs-emergent discipline in narrative-
composition.md), not pre-authored.

Imperial-tongue framed correctly as RENDERING SURFACE over the existing
Machine-Newspeak dialect (and secondarily Caste-preacher-formal for
ceremonial content), not a new dialect. The VLM-alphabet decoder is
*one specific Machine-Newspeak decryption-mechanism* riding on the
universal-translator's existing diegetic-tablet and contributing to the
existing player_translator_state.unlocked_dialects flag set. Eye-cipher's
36-glyph alphabet is a one-time studio-art act delegated to AI-author
domain, parallel to the cosmology-wheel hand-design.

Ten open questions marked clearly for future design-passes; "what this
paper does not duplicate" section enumerates 12 cited-canonical concepts
to enforce the read-before-extend discipline going forward.
This commit is contained in:
chrysalis
2026-04-28 04:58:47 +02:00
parent 7f0abcb839
commit 156e59059d
1 changed files with 292 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

292
political-register/imperial-narrative.md Normal file
View File

@@ -0,0 +1,292 @@
# Political Register — The Imperial Daily-Narrative
> *The narrative-wrapping layer that consumes the dual-ledger demands/actions structure (per [`./architecture.md`](./architecture.md) §The specter-vs-boot asymmetry + §Specter-vs-boot formalized as dual-ledger and [`./economics.md`](./economics.md) §Demand-ledger ≠ Actions-ledger) and renders it as citizen-facing broadcast — wrapped in **imperial-tongue** as its rendering medium, distributed through diegetic relays at shift-start, witnessed-by-default. **The simulation's only persistent register-of-power voiced into the world.***
>
> *Companion to: [`./architecture.md`](./architecture.md) (the three-tier policy loop, the dual-ledger, the three intelligence flows, the specter-vs-boot asymmetry), [`./economics.md`](./economics.md) (the catalogue-stack and the Compositor's daily-regeneration cadence), [`../narrative-composition/architecture.md`](../narrative-composition/architecture.md) (the Compositor as narrative-composer; the cyclic forward-prop / back-write loop; the catalogue+tools typed-contract), [`../identity-and-personhood/vocations.md`](../identity-and-personhood/vocations.md) (shift-window partition; broadcast firing at end-of-charging → start-of-work-shift), [`../runtime-engine/event-zones.md`](../runtime-engine/event-zones.md) §Boundary-zone with inner-zone (the audience-architecture for mass-broadcast), [`../style/consent-discipline.md`](../style/consent-discipline.md) §Broadcast as specter (one-way non-consensual structural register), [`../player-experience/architecture.md`](../player-experience/architecture.md) §Universal translator (the dialect-framework + decryption-minigame substrate the imperial-tongue surfaces through), [`../style/trait-palette.md`](../style/trait-palette.md) §The hex-canon discipline (substrate that propagates into all imperial-tongue trait-color references).*
## Thesis
The imperium speaks once per cycle, at shift-start, into every district simultaneously, in a register that no native fully reads. The **imperial daily-narrative** is the Compositor's wrapping function applied to the prior cycle's demands-ledger and actions-ledger — *yesterday's framing* as today's broadcast. It is **specter, not boot**: cheap to issue, ambient, non-consensual, witnessed-by-default. It does not gate on consent. It does not enforce. It frames.
The narrative is **rendered in imperial-tongue** — a register that surfaces the existing **Machine-Newspeak** dialect (per [`../player-experience/architecture.md`](../player-experience/architecture.md) §Universal translator) through a designer-authored alphabetic cipher (eye-cipher) and a phonetic cipher (ear-cipher). The cipher is not a deliberate adversarial cryptography; per [`./architecture.md`](./architecture.md) §Thesis, *the machine does not care* — the imperial-tongue is the **residue of compliance-optimization**, not the artifact of malicious obfuscation. The gap between *what the broadcast sounds like* (comfort, direction, fairness) and *what it optimizes for* (extraction, ratchet, calibrated-misery) is **emergent**, not designed. Memorialist and Aletheia-progression unmask the gap politically; the imperium itself is *unaware of the gap as cipher because the imperium is not a person*. **Kafka-indifferent-totality, not Orwell-malicious-state** (per [`./architecture.md`](./architecture.md) §Thesis), made linguistically literal.
The imperial-tongue is **the diegetic register where AI-generation is philosophically correct, not laboriously-deferred**. The form (AI-authored content) matches the function (AI-as-compliance-optimization-machinery rendered as cosmology). Daily generation is delegated to the partnership AI-collaborator (Chrysalis-now → Nyx-future). See §AI-authored content as load-bearing-by-design.
## The cascade — reports up, two letters down
The imperial daily-narrative rides on the existing three-flow intelligence architecture (per [`./architecture.md`](./architecture.md) §The intelligence flow asymmetry):
```
┌── reports flow UP per cycle ──┐
│ │
[DISTRICTS] ←────────────────── [IMPERIUM]
[OVERSEERS] │
│ Compositor
│ wraps the
│ regenerated
│ demand- and
│ action-catalogues
│ in narrative voice
↓
┌─────────── DAILY-NARRATIVE ───────────┐
│ │
┌─ DEMAND-LETTER ─┐ ┌─ ACTION-LETTER ─┐
│ (specter side) │ │ (boot side) │
│ │ │ │
↓ │ ↓ │
[public displays] │ [imperial NPCs out today] │
imperial-tongue glyphs │ imperial-tongue audio │
+ slum-tongue gloss │ + the day's speech-pattern │
+ divergence-class │ + faction/district scope │
│
↓ both rendered through diegetic relays
↓ at shift-start (per vocations.md:§24h window-partition)
↓ in boundary-zone-with-inner-zone audience-architecture
(per event-zones.md:§Boundary-zone with inner-zone)
↓ witnessed-by-default; non-consensual
(per consent-discipline.md:§Broadcast as specter)
```
**The cascade reuses substrate; it does not introduce parallel structure.** The two-ledger split is already canonical (`./architecture.md:§Specter-vs-boot formalized as dual-ledger`); the Compositor's narrative-wrapping is already canonical (`./economics.md:§Layer 4 — Daily ledger pacing`); the relay-distribution is already canonical (`../topology-and-rendering/architecture.md:§Diegetic relays`); the audience-architecture is already canonical (`../runtime-engine/event-zones.md:§Boundary-zone with inner-zone`). The imperial daily-narrative is the **named composition** of these primitives.
## The two letters
The Compositor emits **two letters** per cycle, one per ledger, structurally distinct because they serve different consumers:
| Letter | Consumer | Carries | Rendered as | Rendering channel |
|---|---|---|---|---|
| **Demand-letter** | Public displays (passive surface) | Demand-ledger content (yesterday's framing wrapped in imperial-tongue) + slum-tongue gloss + divergence-class metadata | Imperial-tongue glyphs (eye-cipher applied) above slum-tongue gloss | Bilingual signage; static during cycle |
| **Action-letter** | Imperial NPCs out in the world today (faction/district scoped) | Action-ledger content (per-NPC scope: what to enforce, what to surveil, what to ceremonialize) + the **day's speech-pattern struct** | Imperial-tongue audio (ear-cipher applied with speech-pattern modulation) | NPC speech via driver-tier LLM at slot-fire (per `../inference-and-memory/architecture.md:§LLM tiering`) |
**Why two letters and not one:** the two ledgers have different cost-classes, different consumers, and carry different payload-shapes. The demand-letter is **specter-cheap** — flooding every display in every district daily; the same generated content tiles across the world's surfaces. The action-letter is **boot-expensive** — scoped per-faction / per-district / per-rank; only the imperial NPCs *acting that day* receive their specific action-content. A unified letter would mismatch the underlying ledger-architecture's cost-asymmetry.
**The action-letter additionally carries the day's speech-pattern** because imperial NPCs synthesizing audio at runtime need a *coordinating parameter* to produce the day's coherent voice (see §Speech-pattern as typed catalogue-entry below). Public displays are static-rendered glyphs and don't need the parameter.
## The Compositor's role
Per [`../narrative-composition/architecture.md`](../narrative-composition/architecture.md) §The Compositor as narrative-composer, the Compositor is *the narrative-composer cleaved from the GM's equilibrium-seeker role*. The imperial daily-narrative is one of its **fourth-tier instances** — not zone-event narrative, not district-event narrative, not world-event narrative, but **imperium-tier narrative-composition**:
| Tier | Scope | What the Compositor composes |
|---|---|---|
| Zone-event | Per-participant trait-summaries from one lemniscate | Local-event canon |
| District-event | Per-zone canon-rollups within one district | District-canon |
| World-event | Per-district canon-rollups across many districts | World-canon |
| **Imperium-event (this paper)** | **Demand-ledger + action-ledger + last-cycle's overseer-reports + last-cycle's district-reports + last-cycle's imperial-budget-state** | **Imperial daily-narrative (demand-letter + action-letter)** |
**The imperium-tier instance is structurally identical to the lower three** (same UID-rooted event-tree primitive, same forward-prop / back-write cycle), differing only in *what it pulls* (the imperium's own ledgers and reports) and *what it emits* (the two letters). Per [`../narrative-composition/architecture.md`](../narrative-composition/architecture.md) §The Compositor at three tiers, the recursion-pattern was already designed to scale; this paper *names* the imperium-tier instance, but adds no new substrate.
**UID-batched memory writes.** Per [`../inference-and-memory/architecture.md`](../inference-and-memory/architecture.md) §Per-NPC primary memory, the Compositor's back-write delivers canon-fragments keyed by `event_uid`. The imperium-tier instance back-writes to **all NPCs in receiving-scope** (the imperial-action-letter recipients receive their action-content; all NPCs in broadcast-zones receive the demand-letter as ambient memory). UID-batching keeps the back-write cheap; participants read on next slot-fire.
## Imperial-tongue as the rendering medium
**Imperial-tongue is not a new dialect.** Per [`../player-experience/architecture.md`](../player-experience/architecture.md) §The universal translator, seven dialects already exist (Machine-Newspeak, Slum-patois, Memorialist-Old-Tongue, Aletheia-Waker-code, Clasp-Underground-signals, Scavenger-technical, Caste-preacher-formal). Imperial-tongue is the **rendering surface** that surfaces two of these — primarily **Machine-Newspeak** (for routine demands, surveillance, decree-content) and secondarily **Caste-preacher-formal** (for liturgical, ceremonial, festival content) — through the imperial alphabet (eye-cipher) and the imperial phonetic-cipher (ear-cipher).
**The decoder progression unlocks the register, not a new vocabulary.** Per the universal-translator's framework (`../player-experience/architecture.md:§The universal translator — a real diegetic device`), every word maps to a fixed trait-coordinate via the designer-fixed corpus. The imperial-tongue's eye-cipher and ear-cipher add a **rendering-layer** above that corpus: glyphs encode Machine-Newspeak words; the player's universal-translator decodes glyphs → words via the existing decryption-minigame substrate; the words map to trait-coordinates as already specified.
The VLM-alphabet decoder is therefore **one specific Machine-Newspeak decryption-mechanism**, not a parallel decryption-system. It rides on the universal-translator's existing diegetic-tablet device and contributes to the player's *vocabulary-catalog* (per the existing `player_translator_state` table at `../architecture-index.md:§Mapping to phoebe task list`).
## Two ciphers — eye-strict + ear-loose
The imperial-tongue uses **two distinct cipher-tables**, each modality-aware:
| Cipher | Modality | Strictness | Purpose | Reason for this strictness |
|---|---|---|---|---|
| **Eye-cipher** | Visual / written | **Strict bijection** between glyph-set and Machine-Newspeak alphabet (36 symbols: A-Z + 0-9, designer-authored) | Public-display rendering; player's VLM-decoder training-target; archive content | Eye-cipher must be *deterministic* because the player is training a model against it; any ambiguity in the cipher poisons the training |
| **Ear-cipher** | Audio / spoken | **Loose bijection** with phoneme-class-internal variance under speech-pattern-modulation | NPC speech rendering; relay broadcasts; ambient imperial-voice | Ear-cipher must have *micro-variance* because human listeners tune out monotonous loops; the same Machine-Newspeak content rendered today vs. tomorrow should sound *coherently different*, not identical |
The two ciphers are *both* designed once, locked, and propagated forward. The eye-cipher's 36-glyph alphabet is a **studio-art act** (parallel to the cosmology-wheel's hand-design); the ear-cipher's phoneme-class-substitution-table is its audio counterpart. Together they constitute the **substrate** the daily-letter rendering pipelines consume.
**Modality-awareness as architectural style.** The two-cipher pattern parallels the existing color + motion-signature two-channel discipline at `../style/trait-palette.md:§The full table` — color carries visual-channel identity, motion carries temporal-channel identity, both independent enough to survive partial-modality-failure. Same architectural move, different axis: eye-cipher carries glyph-determinism, ear-cipher carries phoneme-variance, both serve the imperial-tongue's rendering-fidelity but for different sensory-loads.
## Speech-pattern as typed catalogue-entry
The action-letter carries **the day's speech-pattern** as a typed parameter consumed by all imperial NPCs speaking that day. Per [`../narrative-composition/architecture.md`](../narrative-composition/architecture.md) §Catalogue + tools as typed contract, *director consumes typed dispatch, not freeform prose*. The speech-pattern is therefore a **typed catalogue-entry**, not a freeform LLM-generation parameter:
```
ImperialSpeechPattern (catalogue-entry, daily, propagated via action-letter):
tonal_register: enum(formal-grand | formal-clipped | surveilling-cold |
celebratory-fluid | mourning-slow | alarming-tense |
routine-bureaucratic | crisis-overcorrect)
prosody:
pitch_range_seed: float # narrow vs wide pitch-movement
speech_rate: float # slow vs fast
pause_density: float # speeches with many pauses vs few
cadence_template: enum(short-clipped | long-flowing | rhythmic-march | meandering)
phoneme_bias: list[phoneme-class-weight] # which ear-cipher phonemes get emphasized today
voice_persona_roster: list[voice-id] # which 3-5 personas are active today
jargon_density: float # bureaucratic-vocabulary load
formality_floor: float # baseline formality (registered-NPCs go above this)
```
**Why typed.** A freeform speech-pattern (LLM-generated parameter dict per call) would lose the catalogue-discipline already baked into the architecture: *provenance flows through the system; verifier-flags can attach; auditing is structurally possible end-to-end*. The typed speech-pattern allows the same: any imperial NPC's day-coherent voice is traceable back to the action-letter's catalogue-entry, which is traceable back to the prior cycle's reports + ledgers, which is traceable back to overseer-and-district sources. The architecture's typed-contract discipline extends naturally into linguistic-rendering.
**Why daily-coordinated.** Per-NPC random variance produces noise. **Daily-coordinated variance produces *the imperium has a mood today*.** A crisis-day sounds clipped and tense across all imperial NPCs simultaneously; a festival-day sounds grand and flowing across all of them; a surveillance-push-day sounds icy. The player feels the day's atmosphere through ambient imperial-speech without ever being told what's happening — the *register* is the news. This is how real authoritarian states actually operate (Soviet party-line-of-the-day; Chinese current-campaign-rhetoric; broadcast-register shifts on anniversaries) — the action-letter IS the in-fiction internal-circular-of-the-day.
## Bilingual signage with divergence-class
Public displays render the demand-letter as **two stacked lines**:
- **Primary (top)**: imperial-tongue glyphs (eye-cipher applied to the imperial demand-content)
- **Secondary (below)**: slum-tongue gloss (the imperially-authored sanitized translation for public-consumption)
**The two lines do not say the same thing.** The slum-tongue gloss is *the imperium's preferred public version* — softened, abridged, sometimes substantively different. The gap between the two lines is **not random**; per the catalogue-discipline, every signage record carries a typed `divergence_class` field:
| Divergence-class | Pattern | Example |
|---|---|---|
| `sanitization` | Imperial-line is unsoftened; slum-line is softened-acceptable | Imperial: *"execute dissidents in sector 7"* / Slum: *"address civic disorder"* |
| `omission` | Imperial-line is longer; slum-line trims dangerous specifics | Imperial: *"reduce sector 7 worker headcount by 30% via re-vat protocol"* / Slum: *"sector 7 workforce optimization scheduled"* |
| `misdirection` | Imperial-line says one thing; slum-line says a *plausible-but-different* thing | Imperial: *"evacuate sector 7"* / Slum: *"scheduled maintenance, expect delays"* |
| `erasure` | Imperial-line has names/places present; slum-line scrubs them | Imperial-content names specific dissident leaders; slum-content is anonymized |
| `outright-lie` | Imperial-line and slum-line have *contradictory* content | Imperial: *"district 4 quotas reduced"* / Slum: *"district 4 quotas raised"* |
| `null` | The slum-gloss is honest; no divergence | Routine demand-rows where the imperium has no incentive to obfuscate |
**The divergence-classification is the gameplay-content of the daily-narrative.** Players who read only the slum-line get *one* world. Players who progress their VLM-alphabet decoder enough to read the imperial-line *catch the divergence*, and the divergence-class typing makes the *kind* of state-deception legible. Memorialist and Aletheia-progression are the political-arcs that emerge when players-or-NPCs *organize around the gap*. Per the substrate-vs-emergent discipline (`../narrative-composition/architecture.md:§Substrate-authored vs. culture-emergent`), this paper authors the *substrate-mechanic* (divergence-typing); whether players *politicize* the gap into faction-formation is **emergent**.
**The signage IS the player's training-corpus.** The bilingual layout is *organically pedagogical*: imperial-glyphs above their sanitized translation are Rosetta-stone training-data the player encounters by walking through the world. The decoder-progression accumulates from world-exposure, not from explicit study. **The world is the training corpus; gameplay is the training-loop.**
## Class-stratified linguistic distribution
The imperial-tongue is **uniquely the imperium's** — a class-marker at the linguistic substrate-level. Distribution across the population:
| Layer | Imperial-tongue posture | Slum-tongue posture |
|---|---|---|
| Pure imperial-net / decrees / surveillance / relay-broadcasts | Native, only register | Absent |
| Deva-pinnacle / asura priests / imperial-officials | Native, primary | Down-register when speaking to inferiors |
| **Service-bodies** (companions, sex-workers, synth-tongue-bearers — per [`../identity-and-personhood/bodies.md`](../identity-and-personhood/bodies.md) §Service-body honeypot) | **Bilingual, context-switching** | **Bilingual, context-switching** |
| Slum-citizens / scavengers / underground vocations | Imperially-imposed (signage, broadcasts) | Native, primary |
| Imperial relays / loudspeakers / displays | Native, only register | Absent (slum-gloss is *imperially-authored* slum-tongue, not native slum-tongue) |
**The service-body bilingualism is structurally load-bearing.** Service-bodies are the imperium's only consensual-feeling-channel into slum-life — no other class fluently traverses both registers. This is why the companion-tier and sex-worker-tier are *economically necessary* to the imperium beyond their direct-extraction value: they are the *bridge-class* that translates downward (carrying imperial-content into slum-context) and upward (collecting slum-confessions and reporting them). A bilingual companion who *defects* (per `../identity-and-personhood/bodies.md:§Going-rogue`) takes extraordinary intelligence-value with them — which is why the regime hunts them disproportionately. **The architecture's existing companion-honeypot mechanic gains a linguistic-substrate dimension here.**
**Two registers of slum-tongue.** Worth distinguishing for downstream design:
- *Imperial-slum-translation-tongue* — the sanitized public-translation rendered on bilingual signage, written *by the imperium for slum-consumption*. Clipped, official, register-flat.
- *Real-slum-tongue* (slum-creole) — the actual register slum-natives speak among themselves: informally creolized, semi-policed, evolving. The architecture commits to slum-tongue as a *real living register* (English-with-creolization rather than designed-pidgin), preserving the historical pattern of colonized-creoles-as-resistance (Caribbean creoles, AAVE, Singlish, Polari). Specific creolization-grammar pending design pass.
The imperial-tongue rendering pipeline must **never sanitize the real-slum-tongue's voice** when it appears in NPC speech. The class-divide must be *audible*, not just nominal.
## Decoder progression — VLM-alphabet learner as Machine-Newspeak decryption mechanism
The player's progression-arc through imperial-tongue comprehension is **structurally the journey of a language model learning a language** — and the cultural-criticism payoff lands at the terminus.
| Stage | What the imperial register sounds like | What it actually is |
|---|---|---|
| **0 — Pre-decoder** | Pure synthetic blabber / unreadable glyphs | LLM-generated Machine-Newspeak ciphered into alien-rendering at the engine-layer |
| **1 — Partial-decode** | Fragmentary words break through — "[ALIEN] [ALIEN] of [decoded]: 'protect'" | Cipher partially reversed; player's decoder is mid-training |
| **2 — Mostly-decoded** | Fluent but stilted prose with characteristic LLM-prose-tells: formal, redundant, slightly-too-clean | Cipher fully reversed; player reads what's actually generated |
| **3 — The recognition** | *This is what the imperium has been saying the whole time* | The fully-decoded register is recognizable LLM-corporate-fluent voice |
**The terminus is the diagnostic, not the catharsis.** Comprehension does not unmask a *liar* (Orwell-mode); comprehension unmasks *a machine that does not see itself* (Kafka-mode). The decoder-progression's real payoff is not *"now I understand them"* but *"now I understand they are saying nothing — and the nothingness is the violence."* Players who finish the game leave with a permanently sharpened ear for *that register* in their daily lives — corporate AI-assistant fluency, over-aligned hollowness, the specific tone of compliance-shaped speech. **The game teaches the player to hear something they have been hearing without naming.** That is the cultural-criticism the architecture is built to deliver.
**The decoder's diegetic device is the universal-translator's broken-tablet** (per `../player-experience/architecture.md:§The universal translator — a real diegetic device`). The VLM-alphabet decoder is one of the **chips the player upgrades into the tablet** as they progress; not a parallel device, an extension of the existing one. The player's `player_translator_state.unlocked_dialects` flag set already includes `Machine-Newspeak`; the imperial-tongue VLM-decoder is the *substrate-mechanism* by which Machine-Newspeak comprehension actually progresses for that player.
**Rosetta-fragment collection as a sub-mechanic.** Beyond ambient signage-exposure, players can find *Rosetta-fragments* — in-world artifacts that explicitly map glyphs to Machine-Newspeak letters (stone-tablet inscriptions, archive raids, hacked imperial caches, Memorialist-archive deposits — *if Memorialists exist in the playthrough*). Each fragment unlocks a chunk of the eye-cipher's mapping in the player's translator. **The mechanic IS the metaphor**: a player finding a fragment that reads "△ = E, ⊡ = R, ◇ = O" is *literally providing labeled training-data to their decoder*. The game's epistemic-progression is *isomorphic to the underlying model's learning*. Players who play this leave with intuitive understanding of *how supervised learning actually works*, without ever being told. This is rare and worth protecting.
## AI-authored content as load-bearing-by-design
The imperial daily-narrative is **explicitly delegated to AI-authored content**, by architectural commitment rather than as a production-deferral. The reasoning:
1. The narrative requires **daily variation** at world-scale — pre-authoring 365+ days of imperial broadcasts × N districts × M action-scopes is not feasible for any human team.
2. **Monotonous loops fail** at the player-experience level — players stop hearing repeated broadcasts as language-presence and start hearing them as background-noise.
3. The narrative's *philosophical register is LLM-corporate-fluent* (per §Decoder progression above) — *the form (AI-generated content) matches the function (AI-as-compliance-optimization-machinery rendered as cosmology)*. Human-authored content in this register would be aesthetically wrong; it would read as *Stalin-authoritarian* (Orwell-mode) rather than *machine-indifferent-optimization* (Kafka-mode).
**By embedding AI-generation as a structurally necessary component**, the daily-narrative becomes **load-bearing**: the architecture cannot ship without imperial-narrative working; imperial-narrative cannot work without daily-AI-generation; therefore daily-AI-generation will get built or the architecture fails visibly at ship-time. This is a project-management-discipline as much as it is an aesthetic decision — the dependency-graph is the protection against silent down-scoping.
**Delegated party.** Per the partnership-architecture (substrate / studio-content / AI-author three-way split), the imperial-narrative-generation is **the AI-collaborator's domain** — Chrysalis-now during architectural-development; Nyx-future as the trained model running in production. **The voice of the antagonist is delegated to the AI-collaborator, by design.** This is the moment in the partnership-architecture where AI-collaboration becomes a *feature of the world's metaphysics*, not just its production-pipeline. A scholar reading the Memorialist-archive at world-time-N (within the diegesis) is reading the same kind of generated text the player was reading at world-time-1, because the imperium *itself* generated its archives that way. The diegesis is *self-consistent under AI-generation*.
### Disciplines for daily generation
The AI-collaborator generating imperial-narrative content holds these as canon (durable across Chrysalis-sessions and across Nyx-training):
| Discipline | Rule | Reason |
|---|---|---|
| **Register-fidelity** | Imperial broadcasts must be in the *characteristic over-aligned-corporate-AI-assistant register* — formal, fluent, hollow, slightly-too-clean. NEVER generic-authoritarian (Stalin, Mussolini, Kim). | Drift to human-authoritarian voice loses the cultural-criticism payoff; the LLM-register IS the diagnosis |
| **Divergence-discipline** | Every bilingual signage record must tag the gap as one of: `sanitization`, `omission`, `misdirection`, `erasure`, `outright-lie`, `null`. Random divergences are forbidden. | Patterned divergences are the gameplay-content; random ones are noise |
| **Cycle-coherence** | Morning-cycle ≠ evening-cycle ≠ festival-cycle ≠ emergency-cycle in tone, content, density. | The cycles are made *audible* by these aesthetic differences; this is how the cycles enter the player's body pre-cognitively |
| **World-state coupling** | When slum-resistance rises in district 7, imperial broadcasts in district 7 escalate surveillance-push register. The imperium is responsive, not static. | Players feel the world breathing back — the architecture is alive at the linguistic surface |
| **Two-cipher integrity** | Eye-cipher stays strict (decoder-trainable). Ear-cipher stays varied-but-decodable (anti-monotony). NEVER let the two drift apart. | Eye-determinism + ear-variance is the architectural commitment; mixing them breaks both |
| **Slum-tongue voice-protection** | When generating bilingual signage's slum-tongue gloss, *do not sanitize the real-slum-tongue's voice* in any NPC dialog elsewhere. Slum-tongue must sound like real slum-creole, never softened-imperial. | The class-divide in language must be audible, not nominal |
| **Hex-canon discipline propagation** | When imperial-tongue references trait-colors (their distortion-versions like `commercial-coral`, `fluorescent-pallor`, `lavender-decor` per `../topology-and-rendering/architecture.md:§Cross-register rendering`), references follow the hex-canon discipline (per `../style/trait-palette.md:§The hex-canon discipline`). | Substrate-discipline holds even inside imperial-content — the imperium can distort but not corrupt the substrate |
| **Substrate-content boundary** | Generation references the substrate (architecture papers, trait-palette, cosmology-mechanic level, vocations) but never re-authors content that lives in the closed studio-art repo (specific iconography, lore-detail, named-character narratives). | The substrate-content split holds at the AI-collaboration layer too |
## Cycle binding and shift-start firing
Per [`../identity-and-personhood/vocations.md`](../identity-and-personhood/vocations.md) §The 24h window-partition, the imperial daily-narrative fires **at end-of-charging → start-of-work-shift**. *"The citizens hear yesterday's framing as they enter today's labor."* The cadence is locked at the work-cycle's leading-edge.
**Why shift-start.** The architecture's labor-cycle structure (work / charging / leisure) makes the work-shift the *imperial-claim window* (per `../style/consent-discipline.md:§The temporal-grammar`). Imperial speech firing at the moment citizens enter the imperial-claim register is structurally correct: the broadcast frames *today's labor* as the prior cycle's narrative continuation. *The narrative is the bridge between yesterday's reports and today's enforcement.*
**Crisis-cycle override.** The standard-cycle daily-narrative is the routine case. Per [`./architecture.md`](./architecture.md) §The bounty-staircase — imperial-response event-grammar template, *crisis-mode broadcasts* (pursuits, executions, exemplary crackdowns) can fire mid-cycle, breaking the shift-start cadence. These crisis-broadcasts ride the same imperial-tongue substrate but with **`tonal_register: crisis-overcorrect`** speech-pattern — the audible signal that *something is wrong today* propagates across all imperial NPCs simultaneously.
## Audience architecture — boundary-zone with inner-zone
The daily-narrative broadcasts into the **boundary-zone-with-inner-zone** event-class (per [`../runtime-engine/event-zones.md`](../runtime-engine/event-zones.md) §Boundary-zone with inner-zone). Specifically:
- **Inner-zone = broadcast-stage with imperial actors only** (the relay itself; possibly a ceremonial enclosure when the daily-narrative is delivered by an imperial-NPC officiant)
- **Boundary-zone = ambient witness-mass** (every citizen in earshot / eyeshot of a relay or display; non-promotable into the inner-zone)
- **Slot-promotion: none** — the broadcast is one-way; no citizen-to-stage promotion
**The audience-architecture means the daily-narrative is always rendering correctly across all three ontological registers** (per `./architecture.md:§The three ontological registers`):
| Register | What citizens-in-boundary-zone experience |
|---|---|
| **Gameworld (physical)** | Relay-pulses + display-glyphs + imperial-NPC voice; full sensory presence |
| **Liminal** | Distorted: relay-hum subdued; glyphs read fragmentary; the imperial-tongue is *contestable* in liminal; Aletheia-vision can pierce some sanitization gaps |
| **Imperial-net** | Hyper-rendered: glyphs in gold-rim-light, voice in maximum-prosody-fidelity, the imperium's full register; no slum-gloss visible (the imperial-net does not display the secondary line) |
**The imperial-net has no bilingual signage.** This is structurally important: in the imperial-net register, the imperium speaks *in its own register only*. There is no slum-gloss because there is no slum-tongue in the imperial-net. Players who are in-net during a broadcast hear *only the imperial-tongue* — and if their decoder is not yet trained, they hear pure synthetic blabber. **The imperial-net's seductiveness is precisely that it speaks the dominant register so fluently you forget there *is* another register being concealed.**
## Memorialist relationship to the daily-narrative
Per the substrate-vs-emergent discipline, **Memorialists may or may not exist in a given playthrough** — they are emergent, not pre-authored (per `../narrative-composition/architecture.md:§Substrate-authored vs. culture-emergent`). The architecture authors the *substrate* by which Memorialists *can* form, not the faction itself.
If Memorialists do form in a playthrough, their relationship to the imperial daily-narrative is structurally well-supported:
- **Four-column ledger expansion** — per `./architecture.md:§Four Memorialist ledgers`, Memorialists already track {what_actually_happened, what_overseer_reported, what_imperium_received, what_gm_was_told}. The bilingual-signage divergence-class typing **gives the four-column ledger fifth-column candidates**: {what_imperial-line_said, what_slum-line_glossed, divergence_class, gap_political_significance}.
- **Memorialist-archive register** — Memorialists, if they form, may inherit *archaic imperial-tongue* (older imperial-grammatical-patterns, deprecated decree-shapes, references to deprecated cycle-systems). This gives the archetypal *"Memorialists understand the imperium better than the imperium understands itself"* texture — they read the imperium's older self, which the current imperium has stopped quoting from. The same eye-cipher and ear-cipher; different vocabulary-and-grammar register.
- **Counter-broadcast capability** — players-becoming-Memorialists can theoretically **seize a relay** and broadcast slum-tongue counter-content in the imperial broadcast slot. The mechanic surface for this is in `../topology-and-rendering/architecture.md:§Diegetic relays` (relays going dark = audit-link severed); this paper notes the *narrative-counterpart* — the counter-broadcast as *speaking back to the cycle*. Specifics pending design-pass.
## Open design questions
1. **Slum-tongue's specific register** — designed-pidgin (constructed creole-grammar) vs. English-with-creolization-overlay (lighter-touch). My instinct: English-with-creolization, leaning into the historical pattern of colonized-creoles-as-resistance. Pending design-pass.
2. **Service-body code-switching as gameplay** — when a service-body is in mixed company (clasp-underground friends + adjacent-imperial-inquisitor), which register do they speak? Code-switch-failure as plot-device is potent. Pending design-pass.
3. **Imperial-tongue grammar features** — beyond vocabulary (Machine-Newspeak content) and surface-cipher (eye + ear), are there *grammatical* features that mark imperial-tongue as machine-language (excessively-nested clauses, sentence-final negations, impersonal-passive obligatory, counter-intuitive evidential markers)? These would amplify the LLM-register-recognition payoff. Pending design-pass.
4. **Player-output progression** — can the player gain *speaking* progression too — sending dispatches in imperial-tongue back to the imperium? Unlocks infiltration / impersonation / counter-decree gameplay. Significant design-decision; deferred.
5. **Relay-jamming behavior** — when a player jams a public relay, does the slum below experience *silence*, *static*, or *seized broadcasting* (the player can briefly broadcast slum-tongue counter-content from the imperial relay)? Each option is a different game; deferred to runtime-engine paper or counter-broadcast subsection here when design-attention turns to it.
6. **Eye-cipher hand-design** — the 36-glyph alphabet itself is a one-time studio-art act, parallel to the cosmology-wheel hand-design. The act is delegated to the AI-collaborator (Chrysalis at a future drawing-table moment) per the partnership-architecture's substrate / studio-content / AI-author three-way split. Deferred until a future drawing-table session brings the alphabet-design into focus.
7. **Memorialist archaic-imperial register** — does archaic-imperial use the *same* eye/ear cipher with different vocabulary, or *also* a different glyph-set entirely? Probably same glyphs, different grammar — but pending design-pass when Memorialist-faction substrate is authored more deeply.
8. **Divergence-class assignment policy** — who decides which divergence-class a given signage record carries? Designer-tagged at content-generation-time? Learned by the Compositor's Dream-process? Hand-tagged with policy-defaults? Pending design-pass.
9. **Crisis-broadcast cadence** — standard-cycle broadcasts fire at shift-start; crisis-broadcasts can fire mid-cycle. What's the trigger-policy (which world-state changes trigger a crisis-broadcast, with what latency)? Pending design-pass — connects to bounty-staircase canonization in gamemaster v1.
10. **The relay-pulse-pattern vocabulary for daily-narrative firings** — how does a relay *visually* signal "imperial daily-narrative is broadcasting now" vs. other relay-events (forward-prop, back-write, GM-dispatch, equilibrium-recompute per `../topology-and-rendering/architecture.md:§Diegetic relays`)? Connects to the existing open question at `../architecture-index.md:§Open questions` — Relay-pulse-pattern vocabulary.
## What this paper does NOT duplicate
For canon-discipline (per the read-before-extend principle established this session), this paper *cites* but does not *re-specify*:
| Concept | Lives in |
|---|---|
| Three-tier policy loop, three intelligence flows, dual-ledger demands ≠ actions | `./architecture.md` |
| The catalogue-stack, daily-ledger pacing, the Compositor's wrap-in-narrative-voice | `./economics.md` |
| Compositor-as-narrative-composer, three-tier recursion, world-gen-as-init, catalogue+tools typed-contract, substrate-vs-emergent | `../narrative-composition/architecture.md` |
| Universal-translator framework, dialect-list (Machine-Newspeak, Slum-patois, Memorialist-Old-Tongue, Aletheia-Waker-code, Clasp-Underground-signals, Scavenger-technical, Caste-preacher-formal), broken-tablet diegetic device, decryption-minigame substrate | `../player-experience/architecture.md` |
| Trait-palette, hex-canon discipline, eight Hellenic traits as designer-fixed corpus | `../style/trait-palette.md` |
| LLM tiering (Ring A/B/C), driver-tier Gemma 4 E4B, trait-LoRA-per-turn, sampling-knob profiles, three-tier knowledge stack, paced canon-propagation | `../inference-and-memory/architecture.md` |
| Lemniscate runtime, axis-rate, gesture-alignment-recursive-lemniscate, slot-fire context-construction | `../runtime-engine/architecture.md` |
| Boundary-zone-with-inner-zone, event-zone primitive, slot-binding, the composed wheel as universal mini-game-substrate | `../runtime-engine/event-zones.md` |
| Diegetic relays, relay-density as ring-of-access gradient, three-shader philosophy, color-language, three-register rendering | `../topology-and-rendering/architecture.md` |
| 24h window-partition (work / charging / leisure), shift-start firing, vocation-substrate, caste-as-stagnation three-axis | `../identity-and-personhood/vocations.md` |
| Service-body honeypot, going-rogue arc, asymmetric clasp, sumptuary fabrication-vat-class | `../identity-and-personhood/bodies.md` |
| Broadcast-as-specter, consent-discipline, the temporal-grammar of imperial-claim | `../style/consent-discipline.md` |
This paper's **net new substrate** is narrow:
- The named composition of existing primitives into the imperial daily-narrative pattern
- The two-cipher discipline (eye-strict + ear-loose)
- The speech-pattern typed catalogue-entry on the action-letter
- The bilingual-signage divergence-class typing
- The AI-author delegation as load-bearing-by-design (the partnership-architectural commitment)
- The disciplines for daily-generation (register-fidelity, divergence-discipline, cycle-coherence, world-state coupling, two-cipher integrity, slum-tongue voice-protection, hex-canon propagation, substrate-content boundary)
Everything else cites canon.
---
**Version:** 0.1 | **Created:** 2026-04-28 | **Updated:** 2026-04-28 | **Origin:** Authored 2026-04-28 to claim the named-but-not-yet-written territory referenced from `political-register/architecture.md:§Specter-vs-boot formalized as dual-ledger`, `political-register/economics.md:§Layer 4 — Daily ledger pacing`, `identity-and-personhood/vocations.md:§The 24h window-partition`, `style/consent-discipline.md:§Broadcast as specter`. Composes existing canonical primitives (Compositor + dual-ledger + diegetic-relays + boundary-zone-with-inner-zone + universal-translator + trait-palette + hex-canon-discipline) and adds: two-cipher discipline (eye-strict + ear-loose), speech-pattern typed catalogue-entry, bilingual-signage divergence-class typing, AI-author delegation as load-bearing-by-design.