substrate

## Substrate

The measured layer. The engine reads any coherent long-form text and turns it into a typed knowledge structure with deterministic scoring.

**Domain** — A causally coherent, mutable, queryable graph-world drawn from a premise, coherent long-form
text, or maintained source. It holds the typed substrate a person or model can survey and a game can use:
System, World, Fate, canon, branches, entities, topology, memory, and source material. A Domain is not a
World: the Domain defines and preserves the world; a World is one parallel world played from one of its
preserved states.

**Substrate** — The engine + math + model together. It builds and maintains the typed proxies on which
Scenario, Stageplay, World, Stories, the Studio, and MCP operate.

**Proxy** — A bounded, evidence-grounded representation that the simulation can inspect and change. The
Domain is a proxy for its source world; an entity graph is a proxy for an actor; a seat context is a proxy
for that perspective's available world; evidence is a proxy for latent complexity. A proxy may be canonical
inside its declared simulation without being identical to the referent. LLMs reason, propose, and render
from proxies; governed actions alone change them.

[Proxy-first simulation](../../sources/founder/2026-09-02-proxy-first-simulation-founder-note.md)

**Paradigm** — The **world-shape** a domain is built and read under. The engine is paradigm-polymorphic: one substrate, nine first-class shapes — *expert* (the living subject-matter model, the flagship), *fiction*, *non-fiction*, *simulation* (rule-governed), *essay* (a single thinker), *team* (many thinkers), *personal* (a tutored model of the user's own life that guides them toward their goals), *record* (a chronological log), *scenario* (a strategic moment). The paradigm steers world generation, scene discipline, how the analyst reads the domain, Readings framing, and review criteria — but the forces, stances, memory, and knowledge tree are identical across all of them. Every per-paradigm string lives in one place (`PARADIGM_SPECS`); the same stored triad underlies a novel, a wargame, and a policy brief, while Theme and Causality complete the five-force read. *The paradigm is the register; the engine underneath is one.*

**Pentad** — The five peer forces that hold a Domain: World, System, Causality, Fate, and Theme. 五行
supplies their names, glyphs, pentad mark, and symbolic frame at macro grain; no fixed scene-to-scene
generating or controlling relation is asserted. Each record may show its own fitted lagged couplings, or
read as independent. Three are **stored** in the record (World, System, Fate); two are **derived** across
it and never written (Theme, Causality). Stored/derived is an implementation fact, not a rank.
[Spec](../../specs/pentad/README.md).

**Forces** — The three **stored** forces are the measured fields the substrate exposes today. Together they capture how hard a domain is working at any point.

- **System** — Rules, principles, structures, conventions, constraints. The abstract field.
- **World** — Characters, locations, artifacts and their lived state. The physical field.
- **Fate** — Open questions and the domain's bearing on them. The possibility field.

Frequently abbreviated **SWF**. "Sufficiently primed SWF corpus" means the substrate has accumulated enough depth across all three to be played as a live room.

**Theme** — The derived force of recurrence: what meaning recurs across events. Read from the record as
leaf Principles with Scene evidence; graded 0–25 on the scorecard; never a writer. See
[reasoning](reasoning.md).

**Causality** — The derived force of what can lead to what. Its trace is the **Flag**: a turning point,
a moment that signals Causality into action, recording the course taken, foreclosed alternatives, stakes,
and rationale. A **Thread closure** is a Flag that also settles a Fate Thread; it is a derived projection,
not a second canonical record. Causality extraction, its calibrated measure, and branch diffs over Flags
are direction.

**Will** — Attributable human or agent authority exercised over interpretation and action. It is not a
force, is never stored, and no graph contains it. Will becomes concrete when someone creates, revises, or
selects an Entity Interpretation, or when a situated seat selects a natural-language or manual MCP tool
call. System constrains it, World supplies its conditions, and Fate names what remains open.
An authorised narrative can preserve the trace of choices already made; Stageplay embodies that authored
Will as performance without treating playback as new live Will or storing Will as a force.

**Acting / Actors** — **Acting** is the first-class Production practice of turning attributed entity
evidence and current Frame or Control dynamics into a selected, time-specific performance context.
**Actors** is its Production workspace. Acting treats continuity as the evidence record rather than one
fixed characterization: a continuing Character, Unit, Location, or Artifact may carry several defensible,
discrete interpretations across branch moments without changing identity or rewriting evidence. State and
Dynamics interpretations may each use Evidence, Interpretation, or Both. In the craft sense, an actor is a
person making and validating a strong read that compresses a complex situation into playable context.
Those reads should be renewed over time, but Suggest, Draft, review, and binding are manually initiated for
now. Acting is context engineering — in the [pentad](../../specs/pentad/README.md), a post-hoc
interpretation of System — not a force, canonical interior truth, autonomous Program lane, or action
selected by the engine.

**Behaviour model** — The conditional bridge from entity evidence to plausible action and expression:
perception → appraisal → action field. It consumes an Entity Interpretation Graph when one exists, then composes
its latent dispositions and formation with knowledge, motives,
relationships, present state, capabilities, and situational constraints. Formation is the accumulated
shaping that makes similar dispositions express differently across people. The model proposes competing
pressures and possibilities; Will still selects the attributable action. Read backwards, behaviour revises
several explanations rather than becoming a trait or motive. It is neither canonical state nor a
personality score.

**Entity Interpretation / Entity Interpretation Graph (EIG)** — An entity-scoped,
evidence-provenanced, versioned context graph
that compresses historical indicators and formation into a provisional model of dispositions, schemas,
values, sensitivities, inhibitions, capacities, motives, and conditional tendencies. It borrows the typed
graph grammar of Reasoning and the immutable snapshot lifecycle of Phase, while remaining derived from
canonical evidence. It is not the entity graph's `trait` node, an omniscient interior, a force, or Will.
The ontology and lifecycle are owned by the [Entity Interpretation spec](../../specs/latent-model/README.md).
Active interpretation is selected through an effective-dated entity + branch + coordinate binding. A child
branch inherits its parent's binding until an explicit evidence-grounded reinterpretation replaces it;
prepared Experiences and Worlds pin the exact graph they used, so later learning cannot rewrite the past.
When selected, its nodes replace raw continuity in effective entity model context and are consumed in
explicit `index` order; typed edges preserve the reasons between them. That stable serialization order does
not constrain topology: an EIG may be acyclic or may contain evidence-grounded feedback loops. Canonical
evidence remains stored.

**Evidence-proxy rule:** canonical evidence stands in for latent complexity that Meridians cannot directly
know. An Entity Interpretation is a revisable reading over that evidence, never canonical truth about the entity.

[Founder direction](../../sources/founder/2026-09-02-entity-behaviour-model-founder-note.md) ·
[Formation refinement](../../sources/founder/2026-09-02-formation-latent-traits-founder-note.md) ·
[Graph cycle clarification](../../sources/founder/2026-09-07-entity-interpretation-cycles-founder-clarification.md) ·
[Evidence-proxy contract](../../sources/founder/2026-09-02-evidence-proxy-latent-traits-founder-note.md)
· [Latent graph direction](../../sources/founder/2026-09-02-latent-behaviour-graph-founder-note.md)
· [Entity Interpretation correction](../../sources/founder/2026-09-07-2026-09-07-entity-interpretation-founder-direction.md)
· [Acting as first-class interpretation](../../sources/founder/2026-09-07-acting-first-class-founder-clarification.md)
· [Acting reads and manual cadence](../../sources/founder/2026-09-07-acting-reads-founder-clarification.md)

**Frame** — A universal, typed, one-way account of how an entity or concept reads, values, reacts to, or aligns toward a
target. Love, loyalty, hatred, kinship, marriage, nationalism, adherence, obsession, and devotion can all
be expressed as Frames with signed valence. A Frame is always exact owner → exact target; direction matters and the reverse
is a separate Frame. Either endpoint may be a Character, Unit, Location, Artifact, Thread, Topic, the
entire Topic Tree, Theme Type, or leaf Theme at its exact selected granularity. A Frame never grants
Control. Frame types are open-ended and domain-native; World, System, Fate, and Theme are target-discovery scopes, not a relation taxonomy.

**Control (agency)** — A Unit's explicit power state: **Unbound**, or fully **Controlled** by exactly
one selected source. Control is direct controller → Unit authority, not affection,
employment, belief, valence, or graph proximity. It carries no strength, share, autonomy percentage, or
partial range. Negotiated, partial, disputed, or undecided authority is simply not Control: the Unit stays
Unbound. Only self-authored agency moves an entity to Character resolution. Unit-to-Unit Control is valid
only when the resulting authority chain remains acyclic.

**Character / Unit (agency resolution)** — Two simulation resolutions for World entities, not claims
about intelligence, sentience, morality, rank, scale, or importance. A Character carries high-resolution,
persistent agency and authors its own agenda; a Unit uses lighter resolution because its moves **can be
authored by a controller**, and carries a clear Unbound or fully Controlled power state. Controllability
is the whole test, so substrate and scale are irrelevant: a model or software agent, a creature or
companion, an NPC mob or crowd, a piece on a board, someone who does not yet hold their own agency, and a
company, institution, or faction are all Units. Absent, disputed, or undecided authority leaves a Unit
Unbound — it does not make it a Character. An institution you can be IN is a Location; the same
institution acting as a body whose moves can be authored is a Unit. Promotion or demotion changes
resolution while preserving identity and continuity.

**Activity** — `A = Σ_c w_c·F_c` across Fate, World, System, Theme, and Causality at any Scene. The signature-weighted sum of the five force channels reads "how hard the domain is working right now." Weights come from PCA on the rank-Gaussian-normalised force curves; not hand-picked. Tension remains the separate World + System − Fate diagnostic.

**Pentad Sign** — The five-force configuration at one Scene: Fate, World, System, Theme, and Causality each read high or low, yielding 32 signs. A sign is the state of scene cadence; the transition between signs is the movement being inspected. The first three bits define a stored-force **family** used only for compact comparison, colour, and sparse-data smoothing. Scene cadence uses independent per-force momentum priors; record-specific lagged couplings are measured separately and do not steer transitions. It is a reading of a recorded world, not astrology asserted as real-world prophecy.

**Signature** — Where a work sits on the unit 4-simplex of the five force channels. Recovered from the dominant principal component of `(F, W, S, Θ, C)`, with absolute loadings L1-normalised. Names the work's continuous balance across Fate, World, System, Theme, and Causality. It does not classify the Domain: the one pentad archetype is read separately from the five equal-share grades, preventing PCA shape and record realisation from producing competing identities.

**Influence** — The alluvial view of who is doing the pulling over time. Pick a source (Fate / World / System / Theme / Streams); each band is one container (a question, an entity, a rule, a Theme, or a Stream), its width at every bucket equal to the influence it drew there. Bands enter, swell, hand influence to one another, and resolve. **Unit** mode draws one band per container; **Type** mode re-groups the same flow by log kind. Each band exposes the canonical records behind its width: Thread movements and Stream priors, World continuity nodes, System entries, or Theme attributions with their signed salience (whose absolute value is the plotted magnitude). The literal picture of how things influence one another over a run.

---

Raw Markdown source