Control plane + system of record for regulated enterprise agents

Consequential agent decisions, governed end to end.

Ontogony puts an explicit authority boundary between what an agent produces and what an organization treats as an official decision.

Ontogony records what the agent knew, what it did, which models it called, which semantic or policy authority judged the result, what was promoted, and what can be replayed or verified offline.

Replay
Fixed synthetic
Evidence
Pre-generated
Customer data
None
Execution
Not a live production system

Proof framing

One decision. Every authority boundary.

The planned replay reserves a place for one fixed synthetic record. Until publication approval, the mechanism is visible and the record is not.

Publication control

Publication gate: closed

The publication set has not been published. Architecture and method remain visible; records, digests, artifacts, and verification outcomes do not.

Inspect the publication gate

Governed loop

Advisory output does not promote itself.

Follow the ordered transfer from source intake to dossier evidence. The labeled review-return edge keeps advisory review distinct from named promotion authority.

Static system map

The governed decision loop

Responsibility moves through seven named stages, then closes through replay and offline verification. Supporting lanes cross the sequence without replacing it.

Governed decision loop topologyEight linked stages run from source intake and lineage through replay and offline verification. Stage eight sends advisory review findings back to source intake without granting promotion authority. Shared mechanics and supporting lanes cross the complete sequence. A text equivalent follows the diagram.Shared mechanics — Ontogony.PlatformCross-cutting mechanicsEvidence canonization · Vendor adapters · Deterministic test range · Workbench projectionCURRENT SUPPORTING LANESTAGE 01Source intakeComponent · MetaboleCURRENTSTAGE 02Temporal contextComponent · AisthesisCURRENTSTAGE 03Governed runComponent · AllagmaCURRENTSTAGE 04Model accessComponent · ConexusCURRENTSTAGE 05Semantic authorityComponent · KanonCURRENTSTAGE 06Promotion authorityComponent · ARGCURRENTSTAGE 07DossierComponent · ALLCURRENTSTAGE 08ReplayMethod · Offline verificationCURRENT
  1. 01Source intake and lineageComponent · MetaboleCurrent

    Lineage-bearing source material

  2. 02Temporal context and reconstructionComponent · AisthesisCurrent

    Reconstructed decision context

  3. 03Governed runComponent · AllagmaCurrent

    Governed operation and event evidence

  4. 04Controlled model accessComponent · ConexusCurrent

    Bounded model result and call record

  5. 05Semantic and policy judgmentComponent · KanonCurrent

    Judgment with cited authority

  6. 06Promotion and officializationComponent · ARGCurrent

    Promoted, denied, or held decision

  7. 07Dossier and verificationComponent · ALLCurrent

    Inspectable dossier boundary

  8. 08Replay and offline verificationMethod · Offline verificationCurrent
Advisory review returns to the loop

Current · Deterministic test range — a supporting lane outside the core decision sequence.

Required handoffAdvisory resultEvidence attachmentAdvisory review returnSupporting laneProhibited ownership

Structured fallback

Topology text equivalent

The same eight-stage sequence shown in the diagram, in reading and keyboard order.

Ordered governed decision loop and responsibility handoffs
StageResponsibilityComponentHandoffRelationship
1CurrentSource intake and lineageMetaboleLineage-bearing source materialRequired handoff
2CurrentTemporal context and reconstructionAisthesisReconstructed decision contextRequired handoff
3CurrentGoverned runAllagmaGoverned operation and event evidenceRequired handoff
4CurrentControlled model accessConexusBounded model result and call recordAdvisory result
5CurrentSemantic and policy judgmentKanonJudgment with cited authorityRequired handoff
6CurrentPromotion and officializationARGPromoted, denied, or held decisionEvidence attachment
7CurrentDossier and verificationALLInspectable dossier boundaryEvidence attachment
8CurrentReplay and offline verificationLoop closureReview findings returned to the loopAdvisory review return
Cross-cutting and supporting lanes
Shared mechanics — Ontogony.Platform
Cross-cutting mechanics used by the governed loop.
Evidence canonization — Flock
A supporting lane for multi-agent evidence handling.
Redacted evidence intake — Vendor adapters
A boundary lane for vendor and runtime evidence.
Deterministic test range — Synthetic Estate
A controlled test lane outside the core decision sequence.
Operator and buyer projection — Workbench
A projection surface, not a source of truth.

Explore the governed loop

Illustrative walkthrough

Watch one synthetic decision cross every boundary.

See what each stage receives, may decide, must not decide, and hands on. The example teaches the loop without presenting a record, evidence, or replay.

Walk the synthetic decision

Inspection path

Proof, bounded outcomes, and domain context

Move from a public proof surface to the responsibility outcome it explains, then inspect where a planned or research domain supplies context. None of these projections is an availability claim.

01 · Proof

Inspect what may be published

Evidence publication pending

Synthetic replay

The planned fixed replay will follow one decision through every authority boundary.

Inspect replay anatomy
Evidence publication pending

Public status

The future ledger will separate implementation, evidence, availability, and readiness claims.

View system status
Evidence publication pending

Contract registry

The future registry will describe guarantees and exclusions at responsibility transfers.

Review the contract method
Evidence publication pending

Offline verification

The standalone verifier checks an approved unpacked bundle without calling private Ontogony services. No approved bundle is currently published.

Review offline verification

02 · Outcomes

Follow responsibilities by intended outcome

Problem → outcome

Governed execution

Problem: External runtimes can perform work without making their authority boundaries or consequential actions inspectable.

Outcome: A run can be explained as an ordered set of bounded operations, gates, and evidence-producing handoffs.

Follow the capability path

Problem → outcome

Semantic and policy judgment

Problem: Model output alone cannot determine which facts, policies, or interpretations are authoritative.

Outcome: A decision record can distinguish advisory output from the authority that judged it.

Follow the capability path

Problem → outcome

Dossier and offline verification

Problem: Reviewers need a durable record and a way to detect whether published files differ from their manifest.

Outcome: The decision loop ends at an inspectable dossier and an independent checksum-verification boundary.

Follow the capability path

Review all capability paths

03 · Domains

Place the loop in a bounded decision context

Planned fixed synthetic demonstration · Evidence pending

Responsible gaming

The fixed launch demonstration is designed around a synthetic withdrawal-risk decision moving through the full governed loop.

Boundary: Technical decision assurance is not a legal or regulatory determination.

Read the domain boundary

Review all domains

Origins

From the problem of becoming to the problem of authority

How can a machine-produced decision become authoritative without becoming opaque?

Ontogony began with a philosophical question about identity, authority, and persistence. The governed loop translates it into an inspectable operational boundary.

Read the origins