Agentic runtime
The software environment that keeps an agentic actor operational across sessions and model changes.It separates persistent identity, authority, runtime state, memory, tools, supervision and model cognition so replacing the model does not silently replace the actor or expand its jurisdiction.
Authority boundary
The explicit limit on what a system, agent, operator or integration is permitted to decide or execute.An authority boundary can include jurisdiction, credentials, target scope, approval requirements, financial limits and escalation rules; technical capability does not expand that boundary by itself.
Deterministic State Engine (DSE)
An engine whose declared output can be reproduced from the same canonical inputs, versioned rules, build and explicit time basis.DSE systems fail closed on required truth that is missing or contradictory instead of inventing a persuasive state.
Activation Integrity Engine (AIE)
The telecommunications DSE implementation for activation readiness, attach, porting, device and provisioning integrity.AIE turns fragmented carrier events into deterministic operating state, correction queues and NOC-facing workflows while preserving the distinction between source events and projected activation state.
BFIE
The financial and evidentiary DSE lane for transaction, account, obligation and reconciled business state.Its purpose is to keep canonical financial calculation inside one engine boundary so dashboards and downstream products do not evolve competing versions of financial truth.
Network-Centric Operations (NCO)
A coordinated operating model in which connected people and systems share governed information, awareness and capability to act effectively across distributed conditions.In the Mediator security architecture, NCO describes how bounded security, evidence, intelligence and command surfaces can operate together. It is not a standalone sibling product, and it does not erase the authority of the participating systems.
Assurance infrastructure
Systems that preserve the evidence needed to inspect what happened across execution, transformation, recovery and release.Assurance infrastructure can include receipts, provenance, manifests, cryptographic checks, custody state, recovery evidence and independent verification, but it does not turn every technical record into a legal or factual conclusion.
Receipt
A structured record of a declared event, action or transition, normally carrying actor, subject, time, operation and relevant identity or content references.A receipt supports reconstruction and accountability; it should be read as evidence of recorded system state, not as magical proof that every upstream assertion was true.
Provenance
The attributable lineage of an observation, artifact, transformation or release: where it came from, which system handled it, what changed and how the current object relates to prior states.Good provenance lets a reviewer distinguish original source, derived material and final release instead of flattening them into one file.
Chain of custody
The documented history of possession, transfer, handling and control for evidence or a governed artifact.Custody complements provenance: provenance explains lineage and transformation, while custody emphasizes who or what controlled the item and when.
Manifest
A declared inventory of the artifacts, versions, identifiers and checks that constitute a packet or release.A manifest makes the release universe explicit so an independent verifier can detect missing, substituted or unexpected material.
Artifact compiler
A governed transformation system that converts approved inputs into validated release-ready outputs such as websites, documents, packets, reports or software artifacts.Compiler-style operation separates input authority, transformation logic, validation and release evidence so the final artifact can be reconstructed from its declared build state.
Replay
The reconstruction of prior or current state from authoritative inputs, versioned configuration and an explicit time basis.Replay is valuable because it tests whether a result follows from the declared state rather than from hidden mutable context.
HOLD
A non-action state used when required truth, approval or consistency is missing.HOLD is an explicit refusal to manufacture state when the engine cannot satisfy its decision contract.
Libre Albedrío
The human-authority gate.A machine rank is a measurement. A score is a measurement. External action requires a named human. The system produces the receipt; the operator owns the decision.
Claim calculus
The public bound on what a record may say.Evidence, measurement, correlation, similarity, anomaly, persistence and hash results stay separate from conclusion, meaning, causation, identity, wrongdoing, authenticity, truth and fraud.
Unknown (⊥)
A marked unknown, unobservable, or unasserted property.⊥ is not zero and not false. It is NotEstablished. Marked unknowns prevent implicit assumption.
Reach Coordination System (RSC)
Signed, idempotent coordination across Reach surfaces so a later operator can reconstruct work from coordination IDs and receipts.VIDERIS carries RSC in the field.
Source truth and derived intelligence
Source truth is the observed or authoritative input as received from its governing source.Derived intelligence is interpretation, ranking, correlation or projection built from that material. Mediator keeps the two distinguishable so a useful inference does not retroactively rewrite the source.
AVSFT
The Mediator persistent-agent runtime protocol: identity, authority, state, memory, tools and model cognition separated into explicit responsibilities.AVSFT is a proper name, not an initialism.
Ghoul
The shared graph-orchestration host that AVSFT agents run on: lifecycle, supervision, recovery and receipt mechanics.A public named system, not slang.
Model Context Protocol (MCP)
The open client-server standard connecting models to data and tools with persistent context; the governed tool-authority plane in production.
Agent-to-agent (A2A)
Protocols letting autonomous agents coordinate across systems, vendors and trust boundaries.
Reach Trace Graph (RTG)
The temporal graph of recorded journeys, handshakes and traversals that Reach preserves.
Federal award
Money the federal government has promised to pay a recipient, obligated as a contract, grant, loan or insurance.Source: USAspending glossary.
Contract (federal)
Agreement between the federal government and a prime recipient to provide goods and services for a fee.Source: USAspending glossary.
Indefinite Delivery Vehicle (IDV)
Vehicle facilitating supply and service orders: BPA, BOA, GWAC, Multi-Agency Contract, IDC, Federal Supply Schedule.Source: USAspending glossary.
Grant
Federal financial assistance to carry out a public project; not repaid.Source: USAspending glossary.
Cooperative Agreement
Assistance grant with extended agency involvement, oversight and reporting.Source: USAspending glossary.
Federal Action Obligation
The dollar amount of federal obligation, de-obligation or liability on an award transaction.Source: USAspending glossary.
Transaction (federal)
An initial award or any amendment or modification to it.Source: USAspending glossary.
Unique Entity Identifier (UEI)
Alphanumeric SAM.gov code uniquely identifying entities registered to do federal business.Source: USAspending glossary.
NAICS
Six-digit North American Industry Classification System code stating the industry a contract record falls in.Source: USAspending glossary.
Product or Service Code (PSC)
Four-character code stating what a contract specifically purchases, distinct from NAICS industry.Source: USAspending glossary.
Independent verification
Inspection of supported properties without needing the originating system’s private execution authority.The verifier should be able to reproduce relevant checks and state the limits of verification, including which questions remain semantic, legal or evidentiary rather than purely technical.