Seithar is the reference implementation of Xenowar doctrine, built by Volund Industries as an integrated cognitive warfare operating system. Where Xenowar provides the doctrinal framework, Seithar provides the operational tooling: a persistent, multi-tenant platform that executes the full cognitive kill chain across human and machine substrates. The system handles cognitive threat detection (Shield), offensive cognitive operations (Sword), and the shared ontological layer that makes both possible.
The architecture reflects a core design principle: cognitive warfare is an ontology problem before it is an algorithm problem. Every capability in the system reads from and writes to a persistent cognitive ontology. Detection without ontology produces uncontextualized alerts. Operations without ontology produce incoherent effects. Seithar treats the ontology as the primary artifact and every other component as a consumer of it.
Each Seithar instance maintains a persistent cognitive ontology that models five entity classes: actors, narratives, networks, threats, and operations. Actors are individual or institutional cognitive agents, whether human or artificial. Narratives are the belief structures and information patterns that propagate through and between actors. Networks are the connection topologies through which narratives travel. Threats are identified adversarial cognitive operations targeting the defended environment. Operations are the friendly cognitive actions planned or in execution.
The ontology is not a static knowledge graph. It is a continuously updated model of the cognitive battlespace. Every sensor input, analyst annotation, and operational outcome feeds back into the ontology. Relationships between entities carry temporal metadata: when a connection was first observed, when it was last active, how its strength has changed over time. This temporal dimension is what separates a cognitive ontology from a conventional knowledge base. The attack surface is not a snapshot. It is a trajectory.
The ontology schema is tenant-specific but structurally consistent across instances. This means that ontology exports from one Seithar deployment can be ingested by another, enabling coalition operations where allied organizations share cognitive threat data without exposing their full operational picture.
Shield is the defensive subsystem. It performs six-signal cognitive threat fusion, integrating six distinct signal types into a unified threat picture: narrative signals (content analysis of information flows), network signals (structural analysis of propagation topologies), behavioral signals (actor activity patterns and anomalies), temporal signals (coordination timing and campaign cadence), technical signals (platform manipulation artifacts, bot signatures, synthetic media indicators), and ontological signals (changes to the cognitive ontology itself that suggest adversarial manipulation).
The sixth signal type is unique to Seithar. Ontological signals detect attacks against the platform's own representation layer. If an adversary attempts to corrupt the ontology through poisoned data injection, Shield identifies the inconsistency between the proposed ontological change and the existing evidential base. This reflexive detection capability is a direct application of the dual-substrate thesis: the platform's own cognitive model is itself a substrate that can be attacked and must be defended.
Sword is the offensive subsystem. It plans and executes cognitive operations using the Seithar Cognitive Techniques (SCT) framework, a standardized library of twelve techniques that map to the seven stages of the Xenowar cognitive kill chain. Each technique has defined input requirements, execution procedures, effect metrics, and ABP collection templates.
The twelve Seithar Cognitive Techniques are:
SCT-001: Narrative Injection. Introduction of a new narrative into a target information environment. SCT-002: Narrative Amplification. Increase in the reach and perceived credibility of an existing narrative. SCT-003: Narrative Suppression. Reduction of a target narrative's visibility and credibility. SCT-004: Ontology Corruption. Degradation of a target's internal representation model to produce systematic misperception. SCT-005: Trust Erosion. Targeted degradation of trust relationships within a network. SCT-006: Trust Fabrication. Construction of synthetic trust relationships to enable subsequent operations.
SCT-007: Decision Steering. Manipulation of a target's decision inputs to produce a specific desired output. SCT-008: Cognitive Denial. Overload or degradation of a target's cognitive processing capacity. SCT-009: Proxy Manipulation. Influence of a target through manipulation of trusted intermediaries rather than direct contact. SCT-010: Temporal Disruption. Manipulation of a target's perception of event timing and urgency to force premature or delayed decisions. SCT-011: Identity Fragmentation. Exploitation of contradictions within a target's self-model to degrade coherent action. SCT-012: Substrate Bridging. Coordinated attack across human and machine substrates simultaneously, exploiting the seam between human and AI cognition in hybrid decision architectures.
SCT-012 is the technique most distinctive to Xenowar doctrine. It has no equivalent in prior PSYOPS or adversarial ML frameworks because it presupposes a target architecture that includes both substrates. Substrate bridging attacks exploit the trust boundary between a human operator and their AI system: corrupting the AI's outputs to manipulate the human, or manipulating the human to override the AI's correct outputs.
Seithar is built for multi-tenant deployment. Each tenant receives a schema-isolated instance with its own cognitive ontology, threat history, and operational data. Isolation is enforced at the database schema level rather than through application-layer access control, ensuring that a compromised tenant cannot access another tenant's cognitive battlespace model. Tenant schemas share compute infrastructure but no data paths.
This architecture supports the operational reality of cognitive warfare: allied organizations need to operate on a shared platform without exposing their ontologies to each other. A coalition Seithar deployment can host multiple national tenants, each with full platform capabilities and complete data isolation, with controlled sharing mediated through explicit ontology export and import at boundaries defined by the tenants themselves.
MiroFish is the simulation substrate integrated into Seithar for cognitive wargaming and operation pre-testing. MiroFish maintains agent-based models of target cognitive environments, populated from the Seithar ontology, that allow operators to simulate the effects of planned cognitive operations before execution. The simulation models both human cognitive responses (using calibrated behavioral models) and AI system responses (using proxy models of target architectures). MiroFish simulation outputs feed back into operational planning as predicted ABP distributions, giving operators quantitative estimates of likely effects.
Seithar exposes 129 MCP (Model Context Protocol) tools for agent orchestration, enabling AI agents to operate within the platform as first-class participants. These tools cover the full operational surface: ontology queries and mutations, Shield alert management, Sword operation planning and execution, MiroFish simulation control, and administrative functions. The MCP interface makes Seithar an AI-native platform. Human operators interact through conventional interfaces. AI agents interact through MCP tools. Both operate on the same ontology and produce the same operational artifacts. This design reflects the dual-substrate thesis at the platform level: the operating system itself treats human and machine operators as equivalent cognitive agents.