Probabilistic systems cannot establish
their own operational authority.
ORYNX externalizes governance authorization
into a deterministic runtime layer.
ORYNX is a universal deterministic governance orchestration runtime for probabilistic systems.
It separates probabilistic generation from deterministic governance orchestration, establishing reproducible authority states before operational effect occurs.
ORYNX does not generate decisions.
ORYNX establishes whether operational effect is admissible.
Input-agnostic. Model-agnostic. Jurisdiction-adaptable. Industry-agnostic.
The Architecture
AI · Agent · Workflow · Model
Probabilistic Output — cannot self-authorize
↓
ORYNX Runtime
Deterministic Operational Authority
ESTABLISHED
NOT_ESTABLISHED
INDETERMINABLE
↓
Operational Effect + Immutable Evidence
CIDRE — append-only · hash-secured · retrievable
ORYNX does not establish truth. ORYNX establishes governance conformity against configured artifacts.
What ORYNX Changes
Authority becomes reproducible
Every output evaluation produces a deterministic, retrievable authority state — before operational effect occurs. Not probability. Not inference. Governance.
Governance becomes observable
When a deterministic governance layer operates consistently, the behaviour of probabilistic systems under stable governance constraints becomes longitudinally measurable.
Decisions become evidentiary
Every governance decision is stored as an immutable CIDRE record — hash-secured, append-only, retrievable. The proof of authorization cannot come from inside the system.
Improvement becomes auditable
Changing Governance Drift Rates against stable frameworks provide auditable evidence of behavioural evolution in governed deployment over time. Not benchmark. Not eval. Runtime evidence.
Behavioural Governance Telemetry (BGT)
ORYNX does not merely evaluate outputs.
ORYNX makes governed probabilistic behaviour observable over time.
BGT is an emergent observational capability of the ORYNX runtime architecture. It does not exist before ORYNX. It cannot exist without deterministic governance states, stable governance frameworks, reproducible authority evaluation, and per-decision CIDRE evidence.
No existing governance system can produce Behavioural Governance Telemetry as defined by ORYNX, because no existing system establishes deterministic governance states over probabilistic runtime behaviour through reproducible authority evaluation and per-decision evidentiary persistence.
Governance Drift Rate (GDR)
A multidirectional longitudinal BGT metric. Stable framework + model change → model impact. New framework → adoption rate. Stable → behavioural baseline.
RLCF — Production Telemetry Signal
Structured corrective output generated per evaluation against an operator-defined framework. Not a training method. Runtime evidence.
Authority Stability
Consistency of authority outcomes per framework over time. Framework Sensitivity: governance variance across framework versions.
What ORYNX Does Not Establish
ORYNX establishes
Governance conformity against configured artifacts
Deterministic, retrievable authority states
Per-decision CIDRE evidence
RLCF corrective signals
Governance Drift Rate over time
ORYNX does not establish
Truth or factual correctness
Risk assessment or liability
Legal conclusions
Model quality or business suitability
Moral or ethical judgement
Three Agnostics — One Deployment
Input-agnostic
Any input. Any model. Chatbot, agent, workflow, rule engine. Any system that produces an output subject to authorization.
Technology-agnostic
Claude, GPT, Mistral, Llama. Any proprietary model. No model changes. No API modifications.
Jurisdiction-adaptable
EU AI Act, MAS MindForge, NIST, FCA, DORA. One architecture. Every regulatory framework.
ORYNX is deterministic governance orchestration infrastructure for systems whose outputs cannot self-establish operational authority. AI systems are currently the most prominent operational class of such systems. They are the first market — not the definition.
Operating Modes
Inline
ORYNX sits in the request/response flow. Evaluates before the output produces operational effect. CIDRE record written before delivery.
Offline
Batch processing or manual review. Retroactive compliance checks. Both modes produce identical CIDRE records.
Roadmap
V0. Complete.
Documentation
Architecture Spec. Regulatory Mapping. ValueMaps.
V1. Live.
Demonstrator
Four regulated use cases. CIDRE audit log. Framework Builder.
V2. First Partner.
Production
Full CIDRE. Framework DB. Full RLCF pipeline.
V3. Cooperative.
Full Agentic
Multi-step workflow authorization. Agentic chain governance.