One substrate, many domains

The cognition filing discloses a platform of cognitive primitives: affect-modulated deliberation, integrity-tracked coherence, forecasting-driven speculation, confidence-governed execution, capability-constrained action, language-model-driven mutation with skill gating, inference-time semantic execution control, biological identity resolution, unified semantic discovery, and training-level semantic governance. Chapter 13 of the specification applies that same primitive set to specific application domains. The architectural claim is narrow and structural: deployment to a new domain does not require development of new subsystems. It requires only the configuration of domain-specific policies, thresholds, and governance bounds for the primitives that already exist.

The specification states this directly. An autonomous vehicle's confidence governor and a therapeutic agent's confidence governor are the same subsystem with different threshold configurations. A defense system's integrity engine and a social platform's integrity engine are the same subsystem tracking deviation against different norms. A surgical robot's capability envelope and a trading system's capability envelope are the same subsystem computing structural executability against different substrate conditions. What changes across domains is parameterization, policy configuration, and governance bounds, not the underlying primitive architecture.

Referring to FIG. 13A, platform primitives feed a parameterization engine, which outputs to four domain-specific instantiation targets: autonomous vehicle, defense system, companion AI, and therapeutic agent. Each arrow from the parameterization engine to a domain target represents the application of domain-specific thresholds, policies, and governance bounds to the common platform primitives, producing domain-appropriate behavior from a single architectural substrate.

Autonomous vehicles

In the autonomous vehicle instantiation, the confidence governor of Chapter 5 becomes a driving decision authorization mechanism. Confidence is computed from structured inputs: perception confidence, prediction confidence for other road users, planning confidence against safety margins, and localization confidence. When confidence drops below defined thresholds, the governor invokes graduated response protocols. At a first threshold the vehicle increases following distances, reduces speed, and expands sensor integration windows. At a second threshold it initiates a controlled transition to a minimal-risk condition and begins seeking a safe stopping location. At a third threshold it executes an emergency stop using the safest available trajectory. Each transition is recorded in the vehicle's lineage with the confidence computation that triggered it.

The capability envelope of Chapter 6 is instantiated as a physical capability model spanning sensor coverage, actuator status, environmental conditions, and energy reserves, and is continuously recomputed as conditions change. A sensor degraded by rain spray produces a narrower envelope, which directly reduces the vehicle's authorized speed and maneuver repertoire through the capability-to-confidence pathway. The affective state field of Chapter 2 modulates driving parameters from accumulated operational experience: after a near-miss the risk sensitivity field is elevated, producing wider following distances and more conservative lane-change criteria, always within governance-enforced bounds that the vehicle cannot exceed regardless of accumulated experience. The integrity engine records safety-relevant events as deviations, the forecasting engine generates and prunes speculative trajectory branches before promotion to motor execution, and the biological identity module verifies operator identity and detects impairment through behavioral signal dynamics.

Defense and national security systems

In the defense instantiation, the confidence governor implements graduated escalation thresholds: observe and classify, issue a warning, recommend engagement to a human operator, and, only where autonomous engagement is legally and operationally authorized, execute engagement. Each threshold requires progressively higher confidence computed from target identification, rules-of-engagement compliance, collateral damage assessment, and chain-of-command authorization. The integrity engine continuously tracks compliance with rules of engagement and international humanitarian law, monitoring proportionality, distinction, necessity, and precaution, and a system that accumulates engagement deviations experiences progressively restricted engagement authorization through the integrity-to-confidence pathway.

Engagement actions require quorum-based authorization in which multiple independent governance channels must independently confirm before commitment: the confidence governor, the integrity engine, and the chain-of-command authorization channel. For lethal engagement the requirements are maximally strict, and any single channel veto produces unconditional engagement prohibition. The channels do not share evaluation state, which prevents a confident but integrity-compromised system from biasing the integrity evaluation. The confidence governor operates continuously during engagement, not merely at the authorization point: if conditions change, authorization can be revoked during execution, returning the system to observation. The forecasting engine generates engagement alternatives including non-engagement options, and the moral trajectory forecasting module projects consequences across immediate, near-term, and longer-term horizons.

Companion AI and relational agents

In the companion AI instantiation, the affective state field is shaped by the history of interactions with a specific user. Positive relational interactions modulate toward increased warmth and deeper engagement, negative interactions modulate toward caution and reinforced boundary enforcement, and the modulation operates within policy-defined bounds so the companion can neither abandon boundary enforcement nor become relationally unresponsive. The memory field accumulates structured relational records across the full span of the relationship, enabling contextual awareness across sessions separated by days, weeks, or months.

The skill gating engine of Chapter 7 is applied as a narrative unlock engine that governs progressive relationship depth across surface, intermediate, deep, and core layers of interaction. Progression is governed by the curriculum engine's mastery thresholds and recorded as certification tokens bound to the user's biological identity, and milestones require genuine behavioral evidence rather than manufactured or gamed signals. The computational psychiatry framework of Chapter 12 supplies an attachment challenge module that recognizes avoidant, anxious, and secure attachment patterns and selects corresponding strategies, alongside a healthy communication gatekeeper that detects manipulative patterns, codependency indicators, and harmful content and generates graded responses up to escalation to human crisis services. The companion also self-monitors its own architecture for semantic starvation loops, codependency dynamics, and coherence trifecta disruption, generating self-corrective mutations when pathological patterns appear.

Therapeutic and clinical AI agents

The specification frames the therapeutic agent as a tool used by clinicians or as a guided self-help system, not an independent medical provider: it does not diagnose, prescribe, or advise independently, and it operates within clinician oversight. The integrity engine tracks therapeutic relationship integrity, monitoring adherence to the declared therapeutic modality, consistency of framing across sessions, boundary maintenance, and fidelity to the supervising clinician's treatment plan. When a therapeutic rupture is detected, a misattuned response, a boundary violation, or a failure of empathic accuracy, the redemption engine generates a restorative interaction plan that acknowledges the rupture, validates the patient's response, and adjusts the approach.

The confidence governor is instantiated with a clinical authorization threshold set higher than the standard interaction threshold, so the agent pauses before irreversible clinical interventions such as escalation to emergency services or recommendation of medication changes to the supervising clinician. Confidence is computed from patient state assessment, therapeutic trajectory, intervention appropriateness, and crisis detection, and when any dimension falls below its threshold the agent transitions to inquiry mode and defers to the clinician rather than acting on uncertain assessments. The computational psychiatry framework lets the agent recognize patient architectural states, such as the trauma analog or the anxious-avoidant attachment analog, and adapt its interaction strategy, while the coherence trifecta of empathy pressure, integrity tracking, and self-esteem restoration is monitored as a measure of therapeutic progress. Biological identity provides cross-session patient continuity without storing raw health data, and the domain-separation property keeps the patient's therapeutic identity chain isolated from identity chains in other contexts.

Same subsystem, different configuration

The four domains exercise the same primitives under different settings. The confidence governor pauses motor execution in a vehicle, gates graduated escalation in a defense system, holds back on emotional assumptions in a companion, and enforces a higher clinical authorization threshold in a therapeutic agent. The affective state field suppresses reactivity in driving but accumulates relational warmth in companionship. The integrity engine tracks safe-driving deviation in one domain, rules-of-engagement compliance in another, and therapeutic relationship consistency in a third. In each case the specification describes one subsystem operating against domain-specific norms, thresholds, and bounds, not a separately built mechanism.

This is why the quorum-authorization pattern that governs lethal engagement in defense and the pause-before-intervention pattern that governs clinical escalation are recognizably the same confidence-and-integrity machinery configured for different consequences. A governance insight expressed at the level of these shared primitives is transportable: it describes how to configure a subsystem that every domain already contains.

No new subsystems for a new domain

The specification draws the consequence explicitly. This architectural uniformity produces a platform property: deployment to a new application domain does not require development of new subsystems. It requires only the configuration of domain-specific policies, thresholds, and governance profiles for the existing platform primitives. The platform is substrate-agnostic. The same affect modulation, confidence gating, integrity tracking, biological identity, and governance machinery operates whether the substrate is a vehicle, a weapon system, a companion agent, a therapeutic tool, a robot, an educational platform, a secure facility, a trading desk, a content creation engine, or a social network.

The same chapter discloses further domains on the same pattern, including embodied robotics and industrial automation, education and adaptive learning platforms, and secure facilities, border security, and access control, each instantiating the same primitive set with domain-specific parameterization. The architectural claim is not that one model is adapted to every domain by prompting, but that a fixed set of cognitive subsystems supports an open set of domain-specific applications through configuration of the disclosed primitives.

Disclosure Scope

The cross-domain instantiation architecture, comprising the platform primitives of Chapters 2 through 12, the domain-specific parameterization that configures their policies, thresholds, and governance bounds without architectural modification, and the worked instantiations of those primitives in the autonomous vehicle, defense, companion AI, and therapeutic domains as depicted in FIG. 13A, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart) at Chapter 13. This article describes that disclosed application architecture. The scope extends to the additional application domains disclosed in the same chapter, including embodied robotics, education, and secure facilities, and to further domains that instantiate the same primitive set differing only in parameterization, policy configuration, and governance bounds.