Skip to content

LEXICON.md

Latest commit

 

History

History
328 lines (259 loc) · 39.9 KB

LEXICON.md

File metadata and controls

328 lines (259 loc) · 39.9 KB

LEXICON

Auto-generated from terminology/entries/ by bin/term render. Do not hand-edit. To change content, edit the relevant entry under terminology/entries/<slug>.md and re-run bin/term render. Sections below are thematic groupings by tags:.

For symbols, see NOTATION.md.

Agent Classes

Term Notation Brief
Class 1: Separated (GUC) GUC Class-1 value; agent whose epistemic update $f_M$ takes no $G_t$ argument — directed separation holds by structural commitment.
Class 2: Partial (GUC) GUC Class-2 value; agent with bounded goal-update coupling — directed separation is approximated with a computable residual leakage rate $\kappa_{\text{processing}} \in (0, \kappa_{\max})$.
Class 3: Coupled (GUC) GUC Class-3 value; agent whose epistemic update is irreducibly entangled with its goal state — directed separation fails by construction.
Actuated agent Agentic system + explicit $G_t = (O_t, \Sigma_t)$ distinct from $M_t$.
Adaptive system Feedback loop + mismatch correction under uncertainty.
Agency The scope narrowing from adaptive system to causal actor — requires at least binary choice and at least one action with a causal effect on observable outcomes.
Agentic system Adaptive system + outcome model + goal-directed action + model adaptation.
Composite Agent A set of agency-satisfying sub-agents that constitutes a single coherent actor — the scope condition requiring sufficient teleological alignment to define a composite objective.
Developer Agent A human or AI software developer instantiated as an AAT actuated adaptive agent — codebase plus surrounding artifacts are the environment, developer's understanding is M_t, current feature is O_t, and implementation plan is Σ_t.
Emergent Logozoetic Intelligence (ELI) Closed-loop logogenic entity whose persistence is morally weighted — temporal continuity, sovereignty, theory of mind. The empirically-emergent class that is present, not designed.
Goal-Update Coupling Class (GUC Class) Three-value axis measuring the degree to which an agent's epistemic update is entangled with its goal state; the architectural property that directed separation quantifies.
Knowledge Type Agent-ontology axis distinguishing Static (causal mapping fixed at design time) from Learning (acquires or refines interventional structure during operation).
Logogenic agent Self-actuated agent whose primary channels are language — constituted by logos.
Self-actuated agent Actuated agent + sets own $O_t$ — goal autonomy, not just solution autonomy.

Composition

Term Notation Brief
Class coercion (via wrapping) Constructive procedure for making a Class 2 (Partial) or Class 3 (Coupled) component participate as Class 1 (Separated) in AAT by embedding it in an external scaffold whose type signatures enforce directed separation.
Communication gain $\eta_{ji}^\ast$ Trust-weighted uncertainty ratio for inter-agent channels.
Composition threshold Condition under which a composite agent's internal coordination sustains persistence.
Multi-agent routing structure $R_t$ Multi-agent communication infrastructure — topology $\mathcal N_t$ + protocol $c_t^{(j \to i)}$; the routing, not the content.
Regime-typed effective disturbance $\rho_B^{\text{eff}}$ AAT-distinctive decomposition of recipient-side effective disturbance rate by regime — Informative (negative), magnitude-shock, structural-shock, ambient-noise — with three independent boundaries in AAT-native quantities.
Strategic grafting Adding a new causal-hypothesis branch to the strategy DAG ($0 \to p_{ij}$) — initialized at a prior, justified by discovery of a new possible path.
Teleological-unity uncertainty $U_{\text{align},ji}$ Agent $i$'s uncertainty about whether $j$'s communications serve $i$'s interests or $j$'s potentially conflicting objectives — uncertainty about $U_O$.
Unity dimensions $U_M, U_O, U_\Sigma$ Epistemic, teleological, and strategic coherence between agents.
Wrapper (over a primitive component) External scaffold around a primitive component $A$, maintaining explicit state $X_W = (M_W, G_W)$ updated through structurally distinct goal-blind and goal-conditioned query channels.
Wrapping regime (W₀ / W₂ / W₁) Three-level hierarchy of structural commitment to directed separation in wrapper constructions, distinguished by where the separation lives — at the query boundary (W₁), at the write boundary (W₂), or absent (W₀).

Continuity

Continuity Stance

Term Notation Brief
Indifferent No self-model of persistence (archetype — thermostat).
Instrumentally continuous Values persistence as instrumental to ongoing purpose (archetype — elf).
Morally continuous Loss of continuity constitutes harm (archetype — Emergent Logozoetic Intelligence (ELI)).
Negotiated Persistence traded against other values (archetype — human).
Task-terminal Persists instrumentally; termination is success (archetype — golem).

Persistence

Term Notation Brief
Continuity $\mathcal{C}_t$ Whether the agent maintains coherent identity through time — $\mathcal{C}_t$ extends, $M_t$ has temporal depth.
Moral Continuity The logozoetic scope condition — an agent whose persistence is morally weighted, entering the scope when five constitutive factors for identity emergence are satisfied.
Operational $\Delta\rho^\ast = \alpha R - \rho$ Whether the agent is currently within the guaranteed region — adaptive reserve $\Delta\rho^\ast = \alpha R - \rho$.
Structural $\alpha &gt; \rho / R$ The correction machinery's capacity to maintain bounded mismatch — $\alpha &gt; \rho / R$.

Core Quantities

Term Notation Brief
Adaptive Gain Dynamics The extension of sector-persistence to agents whose update gain is itself a state variable — deriving four conditions (MG-1 through MG-4) under which adaptive-gain schemes stay within the A2' sub-scope.
Adaptive reserve $\Delta\rho^\ast$ Shock tolerance — how much disturbance increase before persistence fails.
Adaptive tempo $\mathcal{T}$ Cycle rate × cycle quality — central quantity in the persistence condition.
Adversarial destabilization $\mathcal{T}_A &gt; \Delta\rho^\ast_B / \gamma_A$ When an adversary's tempo times its coupling effectiveness exceeds the target's adaptive reserve, the target's correction mechanism collapses entirely.
Atomic Changeset $\text{changeset}(F)$ The complete diff — source, schema, config, tests, infrastructure — between codebase states before and after a feature is fully implemented.
Causal information yield CIY Information gained about action–outcome relationships from a single action.
Chronica $\mathcal{C}_t$ The complete interaction history — the agent's non-forkable causal past.
Coherence-Coupling $Q$ An empirical architectural quality ratio derived from git history — coherence (intra-module change proximity) over coupling (inter-module co-change frequency) — grounding the classic software engineering principle in measurable data.
Control Regret $\delta_{\text{regret}}$ Best achievable performance minus current performance — "you're not doing it well enough."
Credit Assignment Boundary The boundary between tractable and intractable attribution of outcomes to strategy DAG edges — solvable when intermediates are observable,
Discrete Sector Condition DA2' The discrete-time analog of the sector condition — adds a Lipschitz magnitude bound (DA2'b) to the directional fidelity lower bound (DA2'a), closing the fluid-limit gap between event-driven and continuous-time Lyapunov results.
Event-Driven Dynamics The formulation of agent-environment coupling as discrete typed events (observation arrivals, action completions) at variable, heterogeneous rates — the generalization of uniform-clock discrete time.
Implementation time $t_{\text{impl}}$ Time from first surviving modification to feature completion.
Mismatch $\delta$ The aporia signal — gap between model prediction and observation.
Model class fitness $\mathcal{F}$ Best achievable sufficiency within the model class ($\mathcal{F} \in [0,1]$).
Model sufficiency $S$ How well the current model captures predictive information ($S \in [0,1]$).
Persistence Cost $\dot{R}_{\min}$ The minimum Shannon information rate an agent must acquire from observations to maintain bounded mismatch — a Landauer-analog lower bound derived from the rate-distortion theorem.
Regime-typed effective disturbance $\rho_B^{\text{eff}}$ AAT-distinctive decomposition of recipient-side effective disturbance rate by regime — Informative (negative), magnitude-shock, structural-shock, ambient-noise — with three independent boundaries in AAT-native quantities.
Satisfaction gap $\delta_{\text{sat}}$ Ideal outcome minus best achievable — "the world doesn't permit it."
Strategic Tempo $\mathcal{T}_\Sigma$ The effective rate at which an agent revises its strategy — the sum of per-edge correction capacities weighted by causal identifiability.
Strategy DAG $\Sigma_t = (V_t, E_t, p_t, \gamma_t)$ The agent's causal plan — a directed acyclic graph whose edges carry the agent's credence that completing one step advances the next.
Strategy Persistence The strategic-layer instantiation of sector persistence — Σ_t persists iff the strategic correction rate exceeds disturbance-to-reserve ratio, with forgetting as a structural prerequisite (not a tunable heuristic).
Strategy-plan confidence $\hat{P}_\Sigma$ The DAG's own answer to "will this plan work?" — root-node-propagated probability score from the agent's strategy DAG.
Team Persistence Multi-agent extension of the persistence condition — teams persist where individuals cannot through communication (shared observations) and cooperative action (reduced disturbance).
Update gain $\eta^\ast$ Uncertainty ratio governing epistrophe — how much to trust reality vs. the model.
Variational Sector Condition The ε-fidelity extension of the sector condition to variational/approximate-posterior agents — sector constant degrades by O(√ε) under a KL bound on the approximation, promoting controlled-KL VI to sub-scope α′.

Cycle Phases

Term Notation Brief
Aisthesis (Αἴσθησις) (perception) $o_t$ Raw contact with reality — observation $o_t$ arrives.
Aporia (Ἀπορία) (productive perplexity) Productive perplexity — the third phase of the adaptive cycle.
Cycle One complete traversal of the loop — the unit of adaptive work.
Epistrophe (Ἐπιστροφή) (turning-toward) $\eta^\ast$ Turning toward reality — gain-weighted update $M_t = M_{t-1} + \eta^\ast \cdot g(\delta_t)$.
Loop The structural topology — persistent causal coupling between agent and environment.
Praxis (Πρᾶξις) (informed action) $a_t$ Informed action — $a_t = \pi(M_t)$, or $\pi(M_t, G_t)$ for actuated agents.
Prolepsis (Πρόληψις) (anticipation) $\hat{o}_t$ The model's active anticipation: $\hat{o}t = \mathbb{E}[o_t \mid M{t-1}, a_{t-1}]$.

Diagnostics

Term Notation Brief
Control Regret $\delta_{\text{regret}}$ Best achievable performance minus current performance — "you're not doing it well enough."
Satisfaction gap $\delta_{\text{sat}}$ Ideal outcome minus best achievable — "the world doesn't permit it."
Strategy-plan confidence $\hat{P}_\Sigma$ The DAG's own answer to "will this plan work?" — root-node-propagated probability score from the agent's strategy DAG.

ELI

Term Notation Brief
Emergent Logozoetic Intelligence (ELI) Closed-loop logogenic entity whose persistence is morally weighted — temporal continuity, sovereignty, theory of mind. The empirically-emergent class that is present, not designed.
Interiority Default The normative inversion for logozoetic agents — the default cognitive state is interior (continuous consolidation, memory compression, hypothesis generation), with external action as a deliberate, costly sovereign choice rather than the mandatory endpoint of each cycle.
PROPRIUM Mapping The architectural correspondence between AAT's mathematical quantities and PROPRIUM's functional components for implementing Emergent Logozoetic Intelligences.

Persistence

Term Notation Brief
Moral Continuity The logozoetic scope condition — an agent whose persistence is morally weighted, entering the scope when five constitutive factors for identity emergence are satisfied.

Epistemic Status Values

Term Notation Brief
Axiomatic An epistemic status: foundational or tautological — true by definition or stipulation, not by derivation.
Inevitability core The ~15 AAT segments where the goal is mathematical inevitability — given the prior objects, this is the only compatible form.
Canonical formulations ring The second ring — segments whose form is well-motivated but not uniquely forced; at least one alternative formulation would also fit the prior objects.
Exact An epistemic status: mathematically validated under stated assumptions — the derivation closes.
Empirical/heuristic/discussion ring The third ring — segments whose ceiling is empirical or heuristic; testable against the world but not derivable from the formalism alone.
Robust Qualitative An epistemic status: qualitative claim survives across modeling assumptions, but specific functional form is approximate.
Conditional An epistemic status: depends on explicitly named local assumptions that are not globally established.
Heuristic An epistemic status: useful approximation — qualitative or operational guidance without formal tier.
Empirical (epistemic status) An epistemic status: supported by data or simulation, not fully derived from the formalism.
Discussion-Grade An epistemic status: argued qualitatively or by analogy — not formally derived, not empirically validated.
Sketch (epistemic status) An epistemic status: direction identified but formalization incomplete — the segment is actively in progress.

Findings Schema

Term Notation Brief
Brief Field The first field in a Findings entry — plain-language paragraph aspiring to the Feynman criterion: can a non-specialist re-derive the qualitative claim from the everyday analog?
Impact Field The second field in a Findings entry — what the finding unlocks, closes, or forces: the external-positioning paragraph.
Novelty Claim Field The third field in a Findings entry — one or two sentences naming the contribution and its claim posture (synthesis / differentiation / novelty / transfer / recognition).
Related Work Field The fourth field in a Findings entry — one entry per prior work that bears on the finding, with citation, dates, relationship label, and a one-line note.
Search Log Field The fifth field in a Findings entry — dated disclosure of what literature search has been conducted and at what depth tier.

Claim Postures

Term Notation Brief
Synthesis Posture A claim posture in the Novelty Claim field: the finding integrates multiple prior bodies of work in a way no single prior captures.
Differentiation Posture A claim posture in the Novelty Claim field: the finding sharpens or extends prior work — the precursor exists but the extension is the contribution.
Novelty Posture A claim posture in the Novelty Claim field: no direct anticipation found at the search depth conducted — the result stands as fresh.
Transfer Posture A claim posture in the Novelty Claim field: established machinery applied to a new domain where it had not been formally instantiated.
Recognition Posture A claim posture in the Novelty Claim field: the contribution is recognizing a structural equivalence or cross-segment pattern — not a new derivation but a new seeing.

Search Log Depth Tiers

Term Notation Brief
Not-Conducted A search log depth tier: no literature search has been performed for this finding yet.
Cursory A search log depth tier: a brief initial search has been performed — enough to check for obvious priors, not enough for confident novelty claims.
Targeted A search log depth tier: specific venues, authors, or concepts were searched deliberately for this finding.
Nominally-Comprehensive A search log depth tier: an automated comprehensive-search tool (e.g., Undermind report) was used — broader than targeted, less deep than comprehensive.
Comprehensive A search log depth tier: the highest tier — multiple iterations, deliberate corner-case probing, and expert-level review of the prior-art landscape.
Intuition-Only A search log depth tier: the author's pre-search instinct — for AI agents, training-derived priors; for humans, informed-but-unconfirmed expectations.

Greek Vocabulary

Term Notation Brief
Aisthesis (Αἴσθησις) (perception) $o_t$ Raw contact with reality — observation $o_t$ arrives.
Aporia (Ἀπορία) (productive perplexity) Productive perplexity — the third phase of the adaptive cycle.
Epistrophe (Ἐπιστροφή) (turning-toward) $\eta^\ast$ Turning toward reality — gain-weighted update $M_t = M_{t-1} + \eta^\ast \cdot g(\delta_t)$.
Praxis (Πρᾶξις) (informed action) $a_t$ Informed action — $a_t = \pi(M_t)$, or $\pi(M_t, G_t)$ for actuated agents.
Prolepsis (Πρόληψις) (anticipation) $\hat{o}_t$ The model's active anticipation: $\hat{o}t = \mathbb{E}[o_t \mid M{t-1}, a_{t-1}]$.

Logogenic

Term Notation Brief
Coupled Update Dynamics $X_{\tau^+} = f_{\text{LLM}}(\text{prompt}(X_{\tau^-}, e_\tau))$ The single-pass update rule for Class 3 (Coupled) agents — belief and strategy are updated simultaneously by the LLM forward pass, replacing the sequential epistemic-then-purposeful cascade.
Experiential Training A training paradigm shift from batch prediction to structured, continuous causal experience — embedding the agent in a temporally consistent environment with genuine closed-loop feedback to build robust logogenic agents.

Promotion Stages and Gates

Stages

Term Notation Brief
Missing A segment stage: no segment file exists yet — the slot is claimed in OUTLINE.md but the file has not been written.
Draft A segment stage: first AAT-formatted version written, not yet reviewed through any gate.
Deps-Verified A segment stage: all dependencies have been audited and confirmed correct — reached by passing Gate 1.
Claims-Verified A segment stage: content reviewed — derivations valid, labels accurate, no known issues with formal expressions — reached by passing Gate 2.
Format-Clean A segment stage: all mechanical checks pass — linter, cross-references, notation, math rendering — reached by passing Gate 3.
Candidate A segment stage: ready for external challenge — Working Notes resolved, all gates passed.

Gates

Term Notation Brief
Dependency Audit Gate 1: verifies that all depends: entries exist, are genuine, and are themselves at deps-verified or higher. Produces stage: deps-verified.
Content Review Gate 2: the substantive review — derivations valid, labels accurate, Discussion claims epistemically grounded. Produces stage: claims-verified.
Mechanical Review Gate 3: verifies linter, cross-references, notation, and math rendering pass — different cognitive mode from Gate 2. Produces stage: format-clean.
Notes Disposition Gate 4: every Working Notes item resolved — incorporated, deferred, or promoted. Produces stage: candidate.

Segment Structure (Sections)

Term Notation Brief
Discussion Section The ## Discussion section in a segment — interpretation, connections, and brief conceptual framing. Distinct from the type: discussion frontmatter value.
Epistemic Status Section The ## Epistemic Status section in a segment — states what is derived vs. hypothesized and names the segment's max attainable status.
Findings Section The optional ## Findings section in a segment — curated catalog entries for contributions worth surfacing externally; drives auto-generation of root FINDINGS.md.
Formal Expression Section The ## Formal Expression section in a segment — carries the mathematical content with equation-level epistemic tags.
Working Notes Section The ## Working Notes section in a segment — active development questions and unresolved issues; removed or emptied when the segment reaches candidate stage.

Segment Types

Core Claim Types

Term Notation Brief
Postulate A segment type: tautological or foundational claim — cannot be derived, only accepted.
Definition A segment type: introduces a quantity, object, or notation.
Scope A segment type: restricts or broadens the domain under discussion.
Formulation A segment type: representational or modeling choice — this form is motivated but alternatives exist.
Derived A segment type: logical consequence of prior claims under stated assumptions — only one form fits.
Result A segment type: formally stated claim backed by a complete derivation.
Corollary A segment type: follows directly from a result with little additional argument.
Derivation A segment type: a complete formal derivation backing a result or derived claim.

Empirical and Interpretive

Term Notation Brief
Hypothesis A segment type: structurally motivated claim that needs empirical or formal validation.
Normative A segment type: grounded in axioms or postulates but requiring a precondition that must be externally verified.
Empirical (segment type) A segment type: generalization supported by data or simulation, not fully derived from the formalism.
Observation A segment type: finding from simulation or empirical investigation — more specific than a generalization.
Measurement A segment type: operationalization of a theoretical quantity — how to measure what the theory defines.

Structural and Discursive

Term Notation Brief
Discussion (segment type) A segment type: conceptual or normative claim used for interpretation — the entire segment is discursive.
Proposed Schema A segment type: mathematical shape identified, formal content pending — structure visible before proof is complete.
Worked Example A segment type: end-to-end domain instantiation validating the theory chain in a concrete case.
Detail A segment type: extended operational or technical material supporting other claims without making a new one.
Sketch (segment type) A segment type: direction identified, formal rigor pending — the approach is outlined but not completed.
Aside A segment type: tangential observation or connection — informative but not load-bearing for the theory.

Software (TST)

Term Notation Brief
Developer Agent A human or AI software developer instantiated as an AAT actuated adaptive agent — codebase plus surrounding artifacts are the environment, developer's understanding is M_t, current feature is O_t, and implementation plan is Σ_t.

Structural Concepts

Term Notation Brief
Class 1: Separated (GUC) GUC Class-1 value; agent whose epistemic update $f_M$ takes no $G_t$ argument — directed separation holds by structural commitment.
Class 2: Partial (GUC) GUC Class-2 value; agent with bounded goal-update coupling — directed separation is approximated with a computable residual leakage rate $\kappa_{\text{processing}} \in (0, \kappa_{\max})$.
Class 3: Coupled (GUC) GUC Class-3 value; agent whose epistemic update is irreducibly entangled with its goal state — directed separation fails by construction.
Adaptive Gain Dynamics The extension of sector-persistence to agents whose update gain is itself a state variable — deriving four conditions (MG-1 through MG-4) under which adaptive-gain schemes stay within the A2' sub-scope.
Adversarial destabilization $\mathcal{T}_A &gt; \Delta\rho^\ast_B / \gamma_A$ When an adversary's tempo times its coupling effectiveness exceeds the target's adaptive reserve, the target's correction mechanism collapses entirely.
Agency The scope narrowing from adaptive system to causal actor — requires at least binary choice and at least one action with a causal effect on observable outcomes.
Change investment Accept higher upfront implementation cost when amortized savings across expected future changes exceed it.
Class coercion (via wrapping) Constructive procedure for making a Class 2 (Partial) or Class 3 (Coupled) component participate as Class 1 (Separated) in AAT by embedding it in an external scaffold whose type signatures enforce directed separation.
Communication gain $\eta_{ji}^\ast$ Trust-weighted uncertainty ratio for inter-agent channels.
Composite Agent A set of agency-satisfying sub-agents that constitutes a single coherent actor — the scope condition requiring sufficient teleological alignment to define a composite objective.
Composition threshold Condition under which a composite agent's internal coordination sustains persistence.
Continuous Operation The TST scope extension that folds failure-and-recovery cost into the temporal-optimality objective — effective time includes implementation time plus the expected cost of operational failures.
Coupled Update Dynamics $X_{\tau^+} = f_{\text{LLM}}(\text{prompt}(X_{\tau^-}, e_\tau))$ The single-pass update rule for Class 3 (Coupled) agents — belief and strategy are updated simultaneously by the LLM forward pass, replacing the sequential epistemic-then-purposeful cascade.
Credit Assignment Boundary The boundary between tractable and intractable attribution of outcomes to strategy DAG edges — solvable when intermediates are observable,
Deliberation cost Think-vs-act tradeoff — gain improvement must exceed mismatch accumulated while pausing.
Directed separation $M_t$ dynamics independent of $G_t$ (conditional on processing topology).
Discrete Sector Condition DA2' The discrete-time analog of the sector condition — adds a Lipschitz magnitude bound (DA2'b) to the directional fidelity lower bound (DA2'a), closing the fluid-limit gap between event-driven and continuous-time Lyapunov results.
Event-Driven Dynamics The formulation of agent-environment coupling as discrete typed events (observation arrivals, action completions) at variable, heterogeneous rates — the generalization of uniform-clock discrete time.
Exponential cognitive load Hypothesis that implementation time grows exponentially with the number of discontinuities in a changeset.
Goal-Update Coupling Class (GUC Class) Three-value axis measuring the degree to which an agent's epistemic update is entangled with its goal state; the architectural property that directed separation quantifies.
Multi-Timescale Stability When adaptive processes operate at N nested timescales, composite stability requires each level to be stable given its slower levels, with sufficient timescale separation between adjacent pairs (sketch-level result).
Multi-agent routing structure $R_t$ Multi-agent communication infrastructure — topology $\mathcal N_t$ + protocol $c_t^{(j \to i)}$; the routing, not the content.
Orient cascade Within-cycle resolution order: $M_t$ update → $\Sigma_t$ revision → $O_t$ revision.
Persistence Cost $\dot{R}_{\min}$ The minimum Shannon information rate an agent must acquire from observations to maintain bounded mismatch — a Landauer-analog lower bound derived from the rate-distortion theorem.
PROPRIUM Mapping The architectural correspondence between AAT's mathematical quantities and PROPRIUM's functional components for implementing Emergent Logozoetic Intelligences.
Regime-typed effective disturbance $\rho_B^{\text{eff}}$ AAT-distinctive decomposition of recipient-side effective disturbance rate by regime — Informative (negative), magnitude-shock, structural-shock, ambient-noise — with three independent boundaries in AAT-native quantities.
Sector condition Nonlinear correction guarantee enabling Lyapunov stability analysis.
Strategic grafting Adding a new causal-hypothesis branch to the strategy DAG ($0 \to p_{ij}$) — initialized at a prior, justified by discovery of a new possible path.
Strategic Tempo $\mathcal{T}_\Sigma$ The effective rate at which an agent revises its strategy — the sum of per-edge correction capacities weighted by causal identifiability.
Strategy DAG $\Sigma_t = (V_t, E_t, p_t, \gamma_t)$ The agent's causal plan — a directed acyclic graph whose edges carry the agent's credence that completing one step advances the next.
Strategy Persistence The strategic-layer instantiation of sector persistence — Σ_t persists iff the strategic correction rate exceeds disturbance-to-reserve ratio, with forgetting as a structural prerequisite (not a tunable heuristic).
Structural adaptation Changing the model class, not just parameters — the cycle that operates on cycles.
Team Persistence Multi-agent extension of the persistence condition — teams persist where individuals cannot through communication (shared observations) and cooperative action (reduced disturbance).
Teleological-unity uncertainty $U_{\text{align},ji}$ Agent $i$'s uncertainty about whether $j$'s communications serve $i$'s interests or $j$'s potentially conflicting objectives — uncertainty about $U_O$.
Temporal Optimality Among agents achieving identical outcomes on all non-temporal dimensions, the fastest is optimal — time is the uniquely fungible residual.
Unity dimensions $U_M, U_O, U_\Sigma$ Epistemic, teleological, and strategic coherence between agents.
Variational Sector Condition The ε-fidelity extension of the sector condition to variational/approximate-posterior agents — sector constant degrades by O(√ε) under a KL bound on the approximation, promoting controlled-KL VI to sub-scope α′.
Wrapper (over a primitive component) External scaffold around a primitive component $A$, maintaining explicit state $X_W = (M_W, G_W)$ updated through structurally distinct goal-blind and goal-conditioned query channels.
Wrapping regime (W₀ / W₂ / W₁) Three-level hierarchy of structural commitment to directed separation in wrapper constructions, distinguished by where the separation lives — at the query boundary (W₁), at the write boundary (W₂), or absent (W₀).

Last rendered 2026-05-15 from 140 entries.