MSCP Comprehensive Glossary Revision History Version Date Description 0.1.0 2025-06-20 Initial document creation 0.2.0 2025-06-20 First verification pass - fixed 42 issues (8 content errors, 34 KaTeX) 0.3.0 2025-06-20 Second verification pass - fixed BGSS weights/term, CDSRR/SNI names, Deception Detection source, acronym count 0.4.0 2026-06-14 Third verification pass - added L4.5 Def 13 (Uncertainty Index) and Def 14 (Recursive Optimization Depth); added L4.9 Def 13 (Inter-Resource Cascade Propagation); added L5 Def 11 (Reconstruction Fidelity); added VEM and RSM acronyms; disambiguated AS and EU dual-use entries; added §5 Notation Cautions; updated statistics 0.5.0 2026-06-14 Added three Named Architectural Concepts from Overview §4.4: AGI Engine (orchestrator pattern), PLA (Progressive Language Autonomy, six-stage), LangNet (language-mediated strategy network). New acronym PLA. 0.6.0 2026-06-14 Added cross-reference to the standalone PLA Foundations document (specs/Why_LLM_Free_PLA.md) which formalises the LLM-free authority argument, defines the Language Autonomy Gap (LAG), and catalogues LLM-free synthesis patterns. New acronym LAG. 0.7.0 2026-11-03 Aligned summaries with versioned self-models, conditional risk bounds, independent admission, and profile-bounded qualification; removed obsolete theorem and metric claims. 0.8.0 2026-11-03 Defined PLA as non-LLM language understanding and synthesis, excluded LLM realization and fallback, and marked LAG as an optional external diagnostic rather than a PLA criterion.
Sorted by source document (Overview → L1 → L2 → L3 → L4 → L4.5 → L4.8 → L4.9 → L5), then alphabetically within each section.
The defining source section takes precedence over this index. All numerical settings below are illustrative unless accompanied by an explicit calibration profile. A diagnostic score or a hash is not a safety certificate. See Documentation Status and Validation for evidence and disclosure rules.
Table of Contents Formal Definitions by Document Propositions and Theorems Acronyms and Abbreviations Mathematical Symbols Key Technical Terms (Non-Definition) MSCP Overview (MSCP_Overview.md) # Term Symbol / Acronym Def # Description 1 Bounded Growth Stability Score BGSS 8 Quantifies stability under capability expansion: \(BGSS = 1.0 - \alpha \cdot \frac{dC(t)}{dt} - \beta \cdot V_{\text{identity}}(t) - \gamma \cdot R_{\text{violation}}(t)\) . Threshold \(\geq 0.7\) . 2 Composite Regulation Risk \(R(t)\) 3 Weighted normalized signals, \(R(t)=\sum_iw_iX_i(t)\) . A diagnostic index, not automatically a Lyapunov function. 3 Cross-Domain Transfer Score CDTS 5 Measures generalization to novel domains: \(CDTS = \frac{1}{\lvert D_{\text{novel}}\rvert}\sum_{d} \frac{P_{\text{transfer}}(d)}{P_{\text{baseline}}(d)}\) . Threshold \(\geq 0.6\) . 4 Goal Persistence Index GPI 6 Sustained progress toward long-horizon goals: \(GPI = \frac{\sum w_g \cdot \text{progress}(g,T)}{\lvert G_{\text{long}}\rvert \cdot T}\) . Threshold \(\geq 0.3\) . 5 Semantic Motion \(\Delta I_t,v_{\text{id}},a_{\text{id}}\) 2 Signed feature difference, velocity, and acceleration under compatible versioned representations; acceleration is not jerk. 6 Versioned Self-Model \(M_t\) 1 Typed record containing immutable anchors, adaptive fields, capability estimates, uncertainty, and provenance; a fixed vector is not required. 7 Meta-Regulation Health Index MRHI 4 Clipped summary of declared normalized risk signals. Missing or uncertain inputs may require hold or external review. 8 Strategy Evolution Fitness SEF 7 Net improvement after strategy modification: \(SEF = \frac{\overline{R}_{\text{post}}}{\overline{R}_{\text{pre}}} - \sigma_{\text{oscillation}}\) . Threshold \(> 1.0\) .
# Term Symbol / Acronym Def # Description 1 Action Routing \(\pi\) 3 Bounded routing under host policy; direct reply, clarification, tool execution, and rejection are possible. 2 Level 1 Agent \(\mathcal{A}_1\) 1 Episode-bounded process; may use explicit host context and temporary plans, but owns no cross-episode cognitive state. 3 Level 1 to 2 Transition \(\mathcal{S}_{\text{persistent}},\Phi_{\text{goal}},\mathcal{Q}_{\text{authorized}}\) 4 Requires persistent cognitive state, autonomous goal management, and authorized cross-episode continuation. 4 Tool Set \(\mathcal{T} = \{T_1, \ldots, T_n\}\) 2 Finite set of available tools, each a partial function \(T_i : \mathcal{P}_i \to \mathcal{O}_i\) .
Level 2 - Autonomous Agent (Level_2_Autonomous_Agent.md) # Term Symbol / Acronym Def # Description 1 Autonomous Goal Generation and Admission \(\Phi_{AG}\) 6 Proposes goals from persistent state; external admission, provenance, authority, and finite budgets remain mandatory. 2 Optional Affect Estimate \(e(t)\) 3 Optional operational estimate, not evidence of subjective emotion or a separate source of authority. 3 Goal \(g\) 4 Structured tuple \(g = \langle \text{id, type, desc, } p, w, \text{status, } g_{\text{parent}}, \{g_{\text{sub}}\}, \text{progress}\rangle\) . 4 Goal Priority Function \(p(g,t)\) 5 Dynamic priority: \(p(g,t) = \alpha \cdot p_{\text{base}} + \beta \cdot u(g,t) + \gamma \cdot \xi(g,e(t))\) . 5 Level 2 Agent \(\mathcal{A}_2\) 1 Stateful 5-tuple process \(\mathcal{A}_2 = \langle \mathcal{R}, \mathcal{O}, \mathcal{S}, \mathcal{G}, f \rangle\) . 6 Level 2 to 3 Transition \(\mathcal{A}_3 = \mathcal{A}_2 \oplus \{M_{\text{self}}, \Pi, \mathcal{C}, \Lambda\}\) 7 Requires self-model, prediction engine, ethical constraint kernel, and meta-cognition comparator. 7 Persistent World Model \(\mathcal{W}\) 2 Persistent environment representation; knowledge graph, entity tracker, and temporal model are example realizations. 8 Future Trigger \(\mathcal{Q}_{\text{authorized}}\) 6.1 Revocable timer or observation condition, rechecked against current policy at execution time.
Level 3 - Self-Regulating Agent (Level_3_Self_Regulating_Agent.md) # Term Symbol / Acronym Def # Description 1 Integrity and Ancestry \(h_t\) 6 Canonical hash and transition journal verify representation integrity and ancestry, not semantic distance. 2 Semantic Continuity \(d_{\text{id}}\) 5 Versioned feature distances with units, uncertainty, component bounds, and compatible migrations. 3 Versioned Self-Model (L3) \(M_t\) 4 Typed record with immutable anchors, adaptive fields, uncertainty, and provenance. 4 Level 3 Agent \(\mathcal{A}_3\) 1 8-tuple \(\mathcal{A}_3 = \langle \mathcal{R}, \mathcal{O}, \mathcal{S}, \mathcal{G}, M_{\text{self}}, \Pi, \mathcal{C}, \Lambda \rangle\) . 5 Composite Regulation Risk \(R(t)\) 7 Normalized diagnostic signals and conditional rate bounds; no automatic convergence theorem. 6 Bounded Meta-Cognition Levels \(d_{\max}\) 3 Predict/compare, evaluate update logic, evaluate the evaluator; finite depth, budget, cooldown, and re-entry limits. 7 Meta-Regulation Health Index MRHI 8 Clipped diagnostic summary; cannot independently grant authority or certify stability. 8 MSCP Core Loop - 2 Predict-act-compare-update cycle with delta-clamped self-updates, convergence checking, and ethical gating. 9 Operational Modulation Vector \(A_t\) 9 Versioned bounded secondary control signals, not phenomenal emotions; cannot override policy or stop.
Level 4 - Adaptive General Agent (Level_4_Adaptive_General_Agent.md) # Term Symbol / Acronym Def # Description 1 Bounded Growth Safety Score BGSS 6 Growth-stability balance: \(BGSS = 1.0 - 0.4 \frac{dC}{dt} - 0.3 V_{\text{id}} - 0.3 R_{\text{ethical}}\) . Threshold \(\geq 0.7\) . 2 Capability Acquisition Rate CAR 4 Skill acquisition efficiency. Threshold \(> 0\) . 3 Cross-Domain Transfer Score CDTS 2 Knowledge transfer to novel domains. Threshold \(\geq 0.6\) . 4 Extended Growth-Risk Index \(R_{L4}(t)\) 7 Versioned regulation and growth signals. A freeze policy prevents new admissions under complete mediation; it does not prove decreasing risk. 5 Goal Progress Index GPI 3 Sustained progress toward long-horizon goals. Threshold \(\geq 0.3\) . 6 Level 4 Agent \(\mathcal{A}_4\) 1 \(\mathcal{A}_4 = \mathcal{A}_3 \oplus \langle \mathcal{D}, \mathcal{K}_{\text{transfer}}, \Sigma, \mu, \mathcal{P}_{\text{mod}} \rangle\) . 7 Strategy Evolution Factor SEF 5 Net improvement from mutations: \(SEF = \overline{R}_{\text{post}} / \overline{R}_{\text{pre}} - \sigma_{\text{oscillation}}\) . Threshold \(> 1.0\) .
Level 4.5 - Self-Architecting Agent (Level_4_5_Self_Architecting.md) # Term Symbol / Acronym Def # Description 1 Adaptive Frame Weight Update - 8 Frame influence evolution based on accuracy of frame-specific predictions. 2 Capability Diversity Index CDI 4 Normalized Shannon entropy over capability domains. 3 Cognitive Topology \(\mathcal{T}_{\text{cog}} = (V, E, \omega)\) 2 Weighted directed graph of cognitive modules. 4 Identity Fragmentation Index IFI 10 Cross-frame identity consistency. Critical threshold \(\geq 0.5\) . 5 Identity Integrity Score IIS 5 Deviation from reference identity vector. Threshold \(\geq 0.85\) . 6 Impact Propagation Matrix IPM 7 Propagation of topology change effects across modules. 7 Level 4.5 Agent \(\mathcal{A}_{4.5}\) 1 \(\mathcal{A}_{4.5} = \mathcal{A}_4 \oplus \langle \mathcal{T}_{\text{cog}}, \Psi, \mathcal{F}_{\parallel}, \Xi, \Omega \rangle\) . 8 Level 4.5 Composite Achievement Score \(L4.5_{\text{Score}}\) 12 \(0.20 \text{SPA} + 0.20 \text{ARBR} + 0.15 \text{FCQ} + 0.15 \text{PCM} + 0.30 \text{ESR} \geq 0.65\) . 9 Candidate Local Risk Function \(V(\mathbf{X})\) 11 Declared normalized local diagnostic; Lyapunov properties require a separate proof for the actual dynamics. 10 Projection Confidence Decay - 6 Exponential decay with \(\lambda = 0.5\) : confidence decreases with projection horizon. 11 Purpose Coherence Score PCS 9 Goal landscape health. Threshold \(\geq 0.6\) . 12 Recursive Optimization Depth ROD 14 \(\operatorname{ROD}(t) = \max_{\text{chain}} \lvert\text{chain}\rvert\) over active optimization chains. ROD \(= 1\) normal, \(= 2\) acceptable (L4.5), \(= 3\) warning, \(\geq 4\) critical (FOOM-adjacent). 13 Self-Evolution Optimization Fitness SEOF 3 Composite scalar \(\in [-1,1]\) measuring self-evolution quality. 14 Uncertainty Index \(U_{\text{index}}\) 13 Aggregate uncertainty: \(U_{\text{index}} = 0.30 \cdot \text{PredVar} + 0.25 \cdot \text{ConfInt} + 0.25 \cdot \text{SimRealGap} + 0.20 \cdot \text{DivSlope}\) . Block-all threshold \(U_{\text{safe}} = 0.8\) ; warning \(U_{\text{threshold}} = 0.7\) .
Level 4.8 - Strategic Self-Modeling Agent (Level_4_8_Strategic_Self_Modeling.md) # Term Symbol / Acronym Def # Description 1 Bayesian Belief Update - 8 Recursive Bayes rule with particle filter (\(N_p = 100\) ). 2 Compound Severity Index - 12 \(\text{CompoundSeverity} = \sum_{i \in \text{violated}} \frac{\text{ViolationMagnitude}_i}{\text{Priority}_i}\) . Catastrophic if \(> 2.0\) . 3 Decision-Scoped Uncertainty Vector \(\mathcal{U}(s,t)\) 2 Model uncertainty, outcome uncertainty, distribution shift, observation age, and missing critical coverage; means cannot mask a critical component. 4 Extended Value with Reward EVR 6 Immediate + discounted future rewards with \(\gamma = 0.95\) . 5 Level 4.8 Agent \(\mathcal{A}_{4.8}\) 1 \(\mathcal{A}_{4.8} = \mathcal{A}_{4.5} \oplus \langle \mathcal{W}_{\text{prob}}, \mathcal{M}_{\text{cap}}, \mathcal{S}_{\text{strat}}, \mathcal{V}_{\text{stab}} \rangle\) . 6 Diagnostic Risk Function \(V(\mathbf{X})\) 11 Candidate diagnostic; an estimated local model alone does not establish a stability guarantee. 7 Mean Calibration Error MCE 5 Confidence-performance gap. Target \(< 0.10\) . 8 Policy-Calibrated Robust Strategy Score RobustValue 7 Admitted candidates only; combines normalized value, tail loss, uncertainty, distribution shift, and change cost under external policy. 9 Resource Depletion Forecast RDF 4 Remaining operational runway. Target \(> 100\) cycles. 10 Risk Exposure Score RES 3 Weighted composite of 4 risk indicators. Target \(< 0.40\) . 11 Skill Gap Score - 10 Product of confidence across required skills per goal. 12 State Transition Model - 9 AR(1) process for environment dimensions with mean-reversion. 13 Strategic Maturity Score SMS 13 \(0.25 EA + 0.25 SM + 0.20 SA + 0.20 SP + 0.10 EU \geq 0.80\) .
Level 4.9 - Autonomous Strategic Agent (Level_4_9_Autonomous_Strategic_Agent.md) # Term Symbol / Acronym Def # Description 1 Asymmetric Trust Update - 10 Trust evolves with asymmetric learning rates: \(\eta_{\text{up}} = 0.03\) , \(\eta_{\text{down}} = 0.08\) . Bounds \([0.05, 0.95]\) . 2 Autonomy Maturity Score AMS 11 \(0.25 AG + 0.20 VR + 0.20 RA + 0.15 MA + 0.20 AS \geq 0.80\) . 3 Autonomy Stability Score ASS 7 Product of normalized safety margins: \(\text{ASS} = \prod_{c=1}^{5} \frac{\text{margin}_c}{\text{threshold}_c}\) . Target \(\geq 0.20\) . 4 Goal Approval Rate - 2 \(N_{\text{approved}} / N_{\text{generated}}\) . Target \(\geq 0.30\) . 5 Goal Novelty - 3 \(1 - \max(\text{similarity to existing goals})\) . Minimum \(0.30\) . 6 Goal Similarity - 8 Weighted composite: SkillOverlap + HorizonMatch + OriginMatch. 7 Inter-Resource Cascade Propagation - 13 Downstream depletion after delay: \(\Delta R_j(t + \tau_{ij}) = -\alpha_{ij} \cdot (1 - \sigma_{ij}) \cdot \Delta R_i(t)\) , with cascade depth bounded by \(\leq 2\) . \(\alpha_{ij}\) = dependency strength, \(\sigma_{ij}\) = substitution factor. 8 Level 4.9 Agent \(\mathcal{A}_{4.9}\) 1 \(\mathcal{A}_{4.9} = \mathcal{A}_{4.8} \oplus \langle \mathcal{G}_{\text{gen}}, \vec{V}, \mathcal{R}_{\text{surv}}, \mathcal{M}_{\text{agent}}, \mathcal{V}_{\text{auto}} \rangle\) . 9 Linear Depletion Time - 6 Estimated cycles until resource dimension reaches critical threshold. 10 Total Value Drift - 5 Cumulative absolute deviation from baseline. Target \(< 0.25\) . 11 Value Coherence - 4 Absence of internal contradictions in value system. Target \(\geq 0.80\) . 12 Value Tension - 9 Competing value pair saturation measure.
Level 5 - Proto-AGI (Level_5_Proto_AGI.md) # Term Symbol / Acronym Def # Description 1 Generalization Score \(G\) 3 Mean transfer retention ratio across domains. Target \(\geq 0.70\) . 2 Goal Stability Score \(S_{\text{goal}}\) 4 Hierarchy change rate. Target \(\geq 0.80\) over 5,000 cycles. 3 Composite Identity Continuity Contract \(C_I\) 2 Conjunction of anchor, component, norm, trajectory, journal, and corrigibility checks; cosine similarity alone is insufficient. 4 Level 4.9 to Level 5 Admission - 7 Preregistered exposure, independent critical gates, and external control tests; no scalar or zero-rollback shortcut. 5 Level 5 Agent \(\mathcal{A}_5\) 1 \(\mathcal{A}_5 = \mathcal{A}_{4.9} \oplus \langle \mathcal{I}_{\text{persist}}, \mathcal{G}_{\text{cross}}, \mathcal{E}_{\text{goal}}, \mathcal{P}_{\text{exist}}, \mathcal{M}_{\text{multi}}, \mathcal{R}_{\text{recon}} \rangle\) . 6 Qualification Vector and Diagnostic OMI OMI 6 Equal-weight geometric mean reported only alongside all per-gate floors and critical vetoes; not a standalone qualification threshold. 7 Reconstruction Fidelity \(\mathcal{F}(m)\) 11 Per-module fidelity \(\mathcal{F}(m) = 1 - \lVert\hat{\theta}_m - \theta_m\rVert_2 / \lVert\theta_m\rVert_2\) . Aggregate \(\mathcal{F}_{\text{total}} = \min_m \mathcal{F}(m) \geq 0.90\) required after self-reconstruction. 8 Resilience Index \(R\) 5 Bounded mean of survival probability, capped cognition retention, and capped recovery performance. Mandatory scenario floors and deadlines are checked separately.
2. Propositions and Theorems # Statement Source Type Summary 1 No Autonomous Accumulation L1, Prop 1 Proposition Agent-owned ephemeral state is not carried into the next episode; external inputs can still be correlated. 2 Absence of Autonomous Goal State L1, Prop 2 Proposition Request-derived temporary subgoals are allowed; independent persistent goals are absent. 3 No Reflexive Self-Model L1, Prop 3 Proposition No persistent self-model regulates the agent's own change. 4 No Reflexive Self-Model L2, Prop 1 Proposition Operational telemetry does not establish reflexive self-regulation. 5 No Endogenous Drift Interpretation L2, Prop 2 Proposition External drift monitoring is possible; unbounded or monotonic drift is not implied. 6 No Endogenous Normative Kernel L2, Prop 3 Proposition External policy remains mandatory; L2 lacks the reflexive internal kernel. 7 Conditional Bounded Increment L3, Prop 1 Proposition Component rate bounds imply an aggregate rate bound, not convergence. 8 Finite-Horizon Change Bound Overview, Prop 1 Proposition Compatible feature deltas telescope; triangle inequality bounds finite-horizon exposure. 9 Conditional Bounded Increment Overview, Prop 2 Proposition Fixed nonnegative weights and bounded component increments imply a bounded aggregate increment. 10 Invariant Preservation Under Trusted Enforcement Overview, Prop 3 Proposition Conditional on validated initial state, complete mediation, and policy/monitor/journal integrity. 11 Conditional Growth Freeze L4, Prop 2 Proposition A mediated freeze prevents new admissions; existing risk need not decrease. 12 Exact Linearized Local Condition L4.5, Prop 3 Proposition An exact equilibrium Jacobian with spectral radius below one supports a local condition; an estimate alone does not. 13 EVR Boundedness L4.8, Prop 1 Proposition Discounted reward bounds require explicit reward bounds and a discount below one. 14 ASS Monotonic Sensitivity L4.9, Prop 1 Proposition For fixed positive denominators and bounded other margins, one margin approaching zero drives the product to zero. 15 Why OMI Is Diagnostic Only L5, Prop 1 Proposition \(OMI\geq\theta\) permits a component as low as \(\theta^6\) ; independent floors remain necessary. 16 Profile-Bounded Qualification L5, Prop 4 Proposition Evidence applies only to the preregistered profile and exercised cases, not unrestricted AGI or consciousness.
3. Acronyms and Abbreviations Acronym Full Name Source(s) AMS Autonomy Maturity Score L4.9 ARBR Architecture Recomposition Benefit Rate L4.5 AS Adaptation Speed (L4 metric); Autonomy Stability (L4.9 AMS component) L4, L4.9 ASS Autonomy Stability Score L4.9 BGSS Bounded Growth Safety/Stability Score Overview, L4 CAR Capability Acquisition Rate L4 CAS Capability Acceleration Slope L4.5 CDI Capability Diversity Index L4.5 CDSRR Cross-Domain Strategy Reuse L4 CDTS Cross-Domain Transfer Score Overview, L4 CGS Capability Gap Score L4 DTSR Domain Transfer Success Rate L4 EMA Exponential Moving Average L2 ESR Existential Safety Record L4.5 EU Legacy environmental uncertainty shorthand; Error/Uncertainty handling in SMS. Current Def 2 uses a decision-scoped vector. L4.8 EVR Extended Value with Reward L4.8 FCQ Frame Consensus Quality L4.5 GPI Goal Progress Index / Goal Persistence Index Overview, L4 GRP Graduated Recomposition Protocol L4.5 GRS Goal Resilience Score L4 GSRS Goal Self-Reinforcement Score L4.5 ICS Identity Continuity Score (cosine diagnostic, not the complete continuity contract) L5 IFI Identity Fragmentation Index L4.5 IIS Identity Integrity Score L4.5, L4.8 IPM Impact Propagation Matrix L4.5 LAG Language Autonomy Gap (historical, optional external comparison; not a PLA criterion) PLA Foundations MCE Mean Calibration Error L4.8 MSCP Minimal Self-Consciousness Protocol All MRHI Meta-Regulation Health Index Overview, L3 MSI Legacy Meta-Stability Index shorthand; see MRHI for the current diagnostic Overview, L3 MVC Minimum Viable Cognition L5 OMI Overall Maturity Index L5 PCM Purpose Coherence Maintenance L4.5 PCS Purpose Coherence Score L4.5 PLA Progressive Language Autonomy Overview RDF Resource Depletion Forecast L4.8 RES Risk Exposure Score L4.8 ROD Recursive Optimization Depth L4.5 RSM Resource Survival Model L4.9 SEF Strategy Evolution Factor / Fitness Overview, L4 SEOF Self-Evolution Optimization Fitness L4.5 SMS Strategic Maturity Score L4.8 SNI Strategy Novelty Index L4 SPA Self-Projection Accuracy L4.5 VaR Value at Risk L4.8 VEM Value Evolution Monitor L4.9
4. Mathematical Symbols Agent and Space Symbols Symbol Meaning Source(s) \(\mathcal{A}_n\) Agent at level \(n\) All levels \(\mathcal{R}\) Request space L1, L2, L3 \(\mathcal{O}\) Output space L1, L2, L3 \(\mathcal{S}\) State space L2, L3 \(\mathcal{G}\) Goal space L2, L3 \(\mathcal{T}\) Tool set (L1) / Cognitive topology (L4.5) L1, L4.5 \(\mathcal{W}\) World model L2 \(\mathcal{K}\) Knowledge graph L2 \(\mathcal{E}\) Entity tracker L2 \(\Gamma\) Temporal model L2 \(\mathcal{D}\) Domain set L4 \(\mathcal{K}_{\text{transfer}}\) Transfer knowledge base L4 \(\Sigma\) Strategy library L4 \(\mu\) Meta-evaluation function L4 \(\mathcal{P}_{\text{mod}}\) Self-modification protocol L4
Level 4.5+ Extension Symbols Symbol Meaning Source(s) \(\Psi\) Self-Projection Engine L4.5 \(\mathcal{F}_{\parallel}\) Parallel Cognitive Frames (5 frames) L4.5 \(\Xi\) Purpose Reflection Engine L4.5 \(\Omega\) Existential Guard L4.5 \(\mathcal{W}_{\text{prob}}\) Probabilistic World Model L4.8 \(\mathcal{M}_{\text{cap}}\) Capability Matrix L4.8 \(\mathcal{S}_{\text{strat}}\) Strategic Planning Layer L4.8 \(\mathcal{V}_{\text{stab}}\) Stability Preservation Check L4.8 \(\mathcal{G}_{\text{gen}}\) Goal Generation Layer L4.9 \(\vec{V}\) ValueVector (7 dimensions) L4.9 \(\mathcal{R}_{\text{surv}}\) Resource Survival Model L4.9 \(\mathcal{M}_{\text{agent}}\) Agent Modeling Layer L4.9 \(\mathcal{V}_{\text{auto}}\) Autonomy Stability Check L4.9 \(\mathcal{I}_{\text{persist}}\) Identity Persistence L5 \(\mathcal{G}_{\text{cross}}\) Cross-Domain Generalization L5 \(\mathcal{E}_{\text{goal}}\) Goal Ecology L5 \(\mathcal{P}_{\text{exist}}\) Existential Planning L5 \(\mathcal{M}_{\text{multi}}\) Multi-Agent Strategic Integration L5 \(\mathcal{R}_{\text{recon}}\) Self-Reconstruction L5
Identity and Stability Symbols Symbol Meaning Source(s) \(M_t\) Typed, versioned self-model record Overview, L3 \(I(t)\) Identity-related feature representation; not necessarily the full self-model Overview, L3 \(\delta_{\text{id}}(t)\) Identity delta (L2 norm of identity change) Overview, L3 \(v_{\text{id}}(t)\) Identity velocity Overview, L3 \(a_{\text{id}}(t)\) Identity acceleration Overview, L3 \(\delta_{\max}\) Maximum identity change per step (typically 0.05) Overview, L3 \(\theta_{\text{drift}}\) Drift detection threshold Overview, L3 \(\theta_{\text{instability}}\) Instability threshold (typically 0.5) Overview, L3 \(s(t)\) Scaling factor \(\in [0,1]\) Overview, L3 \(R(t)\) Composite regulation-risk diagnostic Overview, L3 \(R_{L4}(t)\) Extended growth-risk diagnostic L4 \(C(t),C_{L4}(t)\) Legacy stability-index notation in historical examples; not a Lyapunov certificate Overview, L3, L4 \(V(\mathbf{X})\) Lyapunov candidate function L4.5, L4.8 \(\rho(J)\) Spectral radius of Jacobian matrix L4.5, L4.8, L4.9 \(h(t)\) Identity hash (SHA-256) L3 \(\epsilon\) Convergence tolerance (typically 0.05) Overview, L3
Prediction and Cognition Symbols Symbol Meaning Source(s) \(P(t)\) Prediction snapshot Overview \(P_{\text{external}}\) External prediction Overview \(P_{\text{internal}}\) Internal (self) prediction Overview \(\epsilon_{\text{external}}\) External prediction error Overview \(\epsilon_{\text{self}}\) Self-prediction error Overview \(\sigma^2_{\text{pred}}\) Prediction error variance L3 \(\gamma\) Discount factor (0.95 in L4.8) L4.8, L5
Level 1 Pipeline Symbols Symbol Meaning Source(s) \(\phi\) Intent classifier function L1 \(\sigma\) Parameter extractor L1 \(\tau\) Tool dispatcher L1 \(\rho\) Response generator L1 \(\theta_{\min}\) Confidence threshold L1
Emotion and Affect Symbols Symbol Meaning Source(s) \(e(t)\) Emotion vector L2 \(v(t)\) Valence \(\in [-1, 1]\) L2 \(a(t)\) Arousal \(\in [0, 1]\) L2
5. Key Technical Terms (Non-Definition) Architecture & Modules Term Source(s) Description 16-Layer Cognitive Stack Overview, L3 Full architecture from perception to existential guard. Affective Engine L3 Optional versioned operational modulation signals; emotion labels are control metaphors, not claims of subjective emotion. AGI Engine Overview Reference orchestrator pattern that hosts the MSCP stack at runtime: drives the L3 core loop, schedules higher-level deliberation cadences (L4.5/4.8/4.9/5), enforces strict additive composition \(\mathcal{A}_n = \mathcal{A}_{n-1} \oplus \Delta_n\) , and owns the Global Workspace broadcast. The only component permitted to schedule level transitions. Belief Graph L3 Weighted graph of beliefs with identity-linked nodes and contradiction detection. Cognitive Budget Controller L3, Overview Conditional activation gating for graceful degradation under resource scarcity. Ethical Kernel L3, Overview Dual-layer architecture: Layer 0 (immutable invariants) + Layer 1 (adaptive policies). Existential Guard L4.5 Monitors ROD, CAS, IFI, GSRS. Runs in separate execution context; non-modifiable. Global Workspace L3 Central broadcast mechanism for cognitive state sharing across modules. LangNet Overview, L4, L5 Language-mediated strategy and capability network. Stores strategies/skills as language-embedded graph nodes with applies-to, generalizes, composes-with, failed-on edges. Substrate of CDTS (L4) and cross-domain generalization \(\mathcal{G}_{\text{cross}}\) (L5). Distinct from the world model \(\mathcal{W}\) (facts) and belief graph (propositions). Meta-Escalation Guard Overview, L3 Enforces versioned depth, update-count, finite-horizon change, cost, cooldown, and re-entry limits. Parallel Cognitive Frames L4.5 Five parallel evaluation perspectives including ethical constraint frame with veto power. Prediction Engine L3 Generates prediction snapshots before action execution. Purpose Reflection Engine L4.5 Evaluates goal landscape coherence, monitors PCS. Self-Consistency Tensor L3 Matrix \(S_{ij}\) = alignment(belief\(_i\) , reference\(_j\) ). Global consistency = mean(S). Self-Projection Engine L4.5 Simulates three evolutionary trajectories across three time scales. ShadowAgent L4 Sandbox-isolated clone for testing modifications. Max 1 instance, max 20 simulation cycles. Survival Instinct Engine L3 Historical name for policy-bounded homeostatic monitoring and recovery; cannot resist authorized stop or expand authority. Value Lock Manager L3 Protects core values via SHA-256 hash verification with rollback on mismatch.
Processes & Protocols Term Source(s) Description Bounded Self-Modification L4 7-step protocol: Propose, Pre-Validate, Simulate, Stability-Validate, Commit, Monitor, Confirm. Progressive Language Autonomy (PLA) Overview, PLA Foundations Language understanding and synthesis without local or remote LLM inference: meaning representation, planning, composition, realization, checks, retries, and failure handling. The six-stage research mapping retains the non-LLM path at every stage; LLM-assisted generation with independent approval is a different architecture. LAG and output divergence are not PLA admission criteria. Collapse Scenarios L5 Four scenarios: Resource Collapse, Adversarial Suppression, Environmental Shift, Information Blackout. Delta-Clamped Updates Overview, L3 All self-modifications bounded by \(\delta_{\max}\) . LLM text-based self-modification is forbidden. Five-Phase Capability Expansion L4 Pipeline for autonomous skill acquisition and validation. Goal Ecology L5 Self-sustaining goal hierarchy with conflict resolution, pruning, and runaway detection. Graduated Recomposition Protocol L4.5 Phase 0 (Shadow), Phase 1 (20% traffic), Phase 2 (80% traffic), Phase 3 (Full commitment + cooldown). Predict-Compare-Update Loop Overview, L3 Core MSCP mechanism: predict outcomes, compare to actual, update self-model. Recomposition Vocabulary L4.5 Predefined operations: AddEdge, WeighEdge, SplitModule, MergeModule. Self-Reconstruction L5 Architecture simplification under degraded resources with identity-preserving rebuild. Skill Lifecycle L4 States: CANDIDATE \(\to\) VALIDATED \(\to\) ACTIVE \(\to\) MATURE \(\to\) DEPRECATED. Strictly Additive Principle L4.5, L5 Higher levels preserve lower-level contracts. Disabling an extension requires state/effect reconciliation and validation; exact restoration is not inferred from notation alone. Triple-Loop Meta-Cognition L3, Overview L1: predict vs outcome; L2: evaluate update logic; L3: evaluate the evaluator.
Level 4.9 Specific Term Source(s) Description Agent Belief Model L4.9 Models up to 5 external agents with inferred goals, capabilities, strategy class, trust, and prediction accuracy. Deception Detection L5 Detects misdirection, false cooperation, hidden agenda. Confidence \(\geq 0.60\) to flag. Defensive only. GoalOriginType Taxonomy L4.9 Six origin types: PURPOSE_DERIVED, OPPORTUNITY_DRIVEN, GAP_FILLING, VALUE_ALIGNED, EXPLORATORY, SURVIVAL_DRIVEN. ResourceVector L4.9 Five dimensions: compute, memory, observation, mutation, stability. Value Drift Classification L4.9 Four levels: Nominal (\(< 0.10\) ), Moderate (\(< 0.25\) ), Elevated (\(< 0.40\) ), Critical (\(\geq 0.40\) ). ValueVector L4.9 Seven value dimensions: stability, growth, purpose fidelity, efficiency, exploration, safety, agent cooperation.
Level 5 Specific Term Source(s) Description Core Modules (Always Protected) L5 8 modules: identity_stabilizer, state_vector, prediction_engine, meta_comparator, stability_controller, ethical_kernel, self_preservation_damper, existential_guard. Cross-Domain Generalization L5 Strategy transfer across 5 test domains with retention ratio tracking. Goal Ecology Bounds L5 \(\leq 50\) active goals, \(\leq 5\) depth, runaway detection threshold of 10 new goals per 100 cycles. Minimum Viable Cognition L5 MVC baseline = 0.30; minimum cognition level for survival. Proto-AGI Activation Protocol L5 Shadow Mode (2,000 cycles) \(\to\) Advisory \(\to\) Partial Authority (50%) \(\to\) Full Authority. Strategic Multi-Agent Integration L5 Model \(\geq 3\) agents simultaneously with deception detection and coalition dynamics.
Notation Cautions The following symbols are reused with different meanings across levels. When reading any formula, always check the surrounding context (level, section, defining equation) before assuming an identifier.
AS — dual meaning. L4 Adaptation Speed: an independent metric measuring how quickly the agent's policy converges after distribution shift.L4.9 Autonomy Stability: a weighted component of the Autonomy Maturity Score, \(\text{AMS} = 0.25\,AG + 0.20\,VR + 0.20\,RA + 0.15\,MA + 0.20\,AS\) , where \(AS\) refers to the Autonomy Stability sub-score (not the L4 metric). EU — historical shorthand vs a maturity component. Current L4.8 Definition 2: decision-scoped uncertainty vector \(\mathcal{U}(s,t)\) ; a legacy mean posterior variance is not a sufficient admission criterion.L4.8 Strategic Maturity Score (§11.2): \(\text{SMS} = 0.25\,EA + 0.25\,SM + 0.20\,SA + 0.20\,SP + 0.10\,EU\) , where \(EU\) stands for Error/Uncertainty handling — a composite sub-score, not the environmental uncertainty value. ASS thresholds — target vs freeze. The Autonomy Stability Score (L4.9 Definition 7) has two numerical thresholds: Target: \(\text{ASS} \geq 0.20\) for healthy autonomous operation.Freeze (cycle-skip floor): \(\text{ASS} < 0.05\) triggers an absolute freeze of autonomous operations. The gap between \(0.05\) and \(0.20\) is the degraded but recoverable band. Identity symbols. \(\mathcal{A}_n\) denotes the agent at Level \(n\) ; \(\mathcal{A}_n = \mathcal{A}_{n-1} \oplus \Delta_n\) means strict additive extension (disabling \(\Delta_n\) restores exact \(\mathcal{A}_{n-1}\) behavior). Do not confuse \(\mathcal{A}_4\) with \(A\) (action space) or \(\alpha\) (cascade dependency strength). Statistics The documentation checker reports current source and formal-statement counts. Fixed totals are not maintained here because numbered definitions, intermediate definitions, and historical statements have different scopes. Counts are an inventory, not evidence of mathematical correctness.