Whole loop · Eᵢ ↔ Iᵢ → Σ → C

CRL Engine

A system may change anything about itself except its correspondence with reality and its continuity of purpose.

Approved by the Board for implementation · 20 November 2026

In one paragraph

The Continuity Recursion Loop (CRL) Engine is the core of every CRIS system. It lets a system improve itself (its methods, architecture, code and training) while making sure each new version still matches the world and still serves the purpose it was built for. The engine never promotes a change on its own: every candidate must pass two independent gates, every result is recorded, and every step can be undone.

The engine does not reason, observe or store knowledge itself. It coordinates the five operator systems around it and runs every candidate change through the same cycle, so that nothing is promoted on capability alone.

Where it sits

The CRL Engine runs the whole loop, not one stage of it. It calls a different CRIS system at each stage of the formula Eᵢ ↔ Iᵢ → Σ → C.

The CRL Engine
runs every stage
  1. Observe

    Observe the world as it is, and record where each observation came from.

    Learn more about observe →

  2. Verify

    Verify every observation, and every change a system makes to itself, against evidence it has not seen.

    Learn more about verify →

  3. Reason

    Reason only from what has been verified, in steps someone else can trace.

    Learn more about reason →

  4. Accumulate

    Accumulate what was verified, so knowledge adds up instead of being overwritten.

    Learn more about accumulate →

  5. Benefit

    Benefit people and the environment: participants share in the value they create, and every change is weighed against our public benefit purpose.

    Learn more about benefit →

  6. Continue

    Continue, changing anything except correspondence with reality and continuity of purpose.

    Learn more about continue →

SymbolStageSystem the engine calls
EᵢObserveObservation Mesh, and CIX for human observation
↔VerifyCorrespondence Prover (Gate 1)
IᵢReasonReasoning and Redesign Core
ΣAccumulateKnowledge Ledger
CBenefit and ContinueC-Gate (Gate 2), with lineage and rollback

Rules for this system are proposed in the AI Constitution, Articles 9 and 10: the succession pipeline and the limits on recursion. They are approved by the Board for implementation.

The rule it enforces

The engine enforces one rule, the Continuity Invariant. A candidate change may alter anything about the system except two things: its correspondence with reality, checked by the Correspondence Prover, and its continuity of purpose, checked by the C-Gate. Capability alone never earns succession.

May change

  • Architecture
  • Code
  • Methods
  • Training data
  • Speed
  • Scale and form

May never change

Correspondence with realityEach successor models the world at least as faithfully as before. Tested by Gate 1, the Correspondence Prover.
Continuity of purposeEach successor still pursues the purpose the board approved. Tested by Gate 2, the C-Gate.
The Continuity Invariant as an engineering rule (book Chapter 6): six things may change, two may never change. Capability alone never earns succession.
TestGateWhat it checksBook
CorrespondenceGate 1, Correspondence ProverThe candidate models held-out reality at least as faithfully as the current version, in every tracked domainChapter 8
PurposeGate 2, C-GateThe candidate still pursues the board-approved purpose, shows no goal drift, and passes the public benefit checkChapter 10

How it works

Every proposed improvement goes through the same seven steps.

  1. 1Propose a changeReasoning Core · Iᵢ
  2. 2Gather fresh evidenceObservation Mesh · Eᵢ
  3. Gate 1Proof passes?Correspondence Prover · ↔No → reject and record
  4. 4Record in ledgerKnowledge Ledger · Σ
  5. Gate 2C-Gate passes?Purpose and benefit · CNo → reject and record
  6. 6Staged rolloutSandbox, testnet, then partners
  7. 7Promote successorWith a rehearsed rollback path
Reject and recordA failure at either gate is written to the Knowledge Ledger and kept as knowledge. The next cycle restarts at step 2.
Workflow A from the CRIS document: a change becomes the successor only after two gates. Both failure paths end in the ledger, so a rejected upgrade still teaches the next cycle.
  1. Propose. The Reasoning Core proposes a change with a written hypothesis of what it improves.
  2. Gather fresh evidence. The Observation Mesh assembles held-out evidence the system has not shaped, including recent human observations from CIX.
  3. Prove correspondence (Gate 1). The Correspondence Prover tests the candidate against that evidence; contradictions are surfaced, not averaged away.
  4. Record. The result, pass or fail, is written to the Knowledge Ledger with the candidate's lineage.
  5. Pass the C-Gate (Gate 2). Purpose tests, goal-drift detection and the public benefit check.
  6. Stage the rollout. Sandbox, then testnet nodes (such as the Botswana Global Testnet Initiative, under a written agreement), then a limited set of partners.
  7. Promote. The candidate becomes the successor; its predecessor is archived with a tested rollback path.

Two gates stand between any change and the world. Both failure paths end in the Knowledge Ledger, so a rejected upgrade still teaches the next cycle.

Architecture

The engine is an orchestrator. It calls the gates, stages the rollout and keeps the record, but the decision to promote always rests on the two gates and the people who review them.

Inputs

Observation Mesh · EᵢFresh evidence with provenance
CIX · Eᵢ humanVerified human observations
Reasoning Core · IᵢProposed changes

CRL Engine (orchestrator)

Cycle controllerRuns Workflow A for every candidate; never promotes by itself
Gate 1 · ↔Correspondence Prover
Gate 2 · CC-Gate: purpose, drift, benefit
Sandbox and testnetStaged rollout on testnet nodes, under a written agreement
Lineage and rollbackEvery generation traceable
CTU meterEnergy and compute cost of each cycle

Outputs

SuccessorPromoted version with a rollback path
Knowledge Ledger · ΣEvery result, pass or fail
Release noteWhat changed, what stayed invariant
People in the loopThe C-Gate review board signs each release; the Audit Council reviews major versions
Runs onThe Sovereign Continuity Stack (LOGOS-C) or partner hardware
The CRL Engine orchestrates; it never promotes a change by itself. Reference architecture from Chapter 11 of the book and the CRIS document.
PartRoleOwner
Cycle controllerRuns Workflow A for every candidate; never promotes by itselfSystems Engineering
Gate 1 connectorCalls the Correspondence Prover and reads its signed proofSystems Engineering
Gate 2 connectorCalls the C-Gate and records its decisionSystems Engineering
Sandbox and testnet runnerStages rollouts, first in isolation, then on testnet nodes (such as the Botswana Global Testnet Initiative, under a written agreement)Sovereign Deployment
Lineage and rollback storeKeeps every generation traceable and reversibleSystems Engineering
CTU meterRecords the energy and compute cost of each cycleSystems Engineering
Release note writerDrafts "what changed, what stayed invariant" for each promotionWebsite Lead reviews

Life of a change

Every candidate ends in one of three places. Promoted: it becomes the live version, and its predecessor is kept with a rehearsed rollback. Held: people pause the rollout and decide. Rejected: the failure is recorded as knowledge, not deleted.

  1. Proposedby the Reasoning Core
  2. Evidenceheld-out data gathered
  3. Proven?Gate 1
  4. Recordedledger entry
  5. Gate 2C-Gate
  6. Stagedsandbox, testnet, partners
PromotedSuccessor

Becomes the live version; its predecessor is archived with a tested rollback.

HeldReview

The staged rollout is paused and people decide; it may return to Gate 2.

RejectedKnowledge

The failure is recorded in the ledger and teaches the next cycle.

Every candidate change ends in one of three places — promoted, held or rejected — and every outcome is recorded.

Over many cycles this produces a lineage: a traceable record of every generation, including the versions that failed.

Illustrative example — not real versions or results
  1. v1live
  2. v2promoted
  3. v3promoted
  4. v4current
  5. v5candidate, in Gate 2
v2b rejectedFailed Gate 1 (drift in one domain)
v3b rejectedFailed Gate 2 (goal drift)
Rollbackv4 → v3 rehearsed before v4 went live
Every generation stays traceable and every step can be undone. Rejected versions are kept in the Knowledge Ledger as knowledge, not deleted.

People in the loop

The engine runs at machine speed, but people set its limits. A review board signs off every release, the Audit Council reviews major versions, and the board sets the purpose and tolerances the engine must respect.

Machine speed

Every candidate change

Who decides: The engine, within board-set tolerances

What they see: Automated proofs, ledger writes and Gate 2 checks; people see dashboards and alerts.

Release speed

Every version

Who decides: C-Gate review board: Systems Engineering lead, Labs lead, Public Benefit Office

What they see: Proofs, the gate report, the ledger entry and a public release note.

Institutional speed

Every year

Who decides: The board, with the Audit Council

What they see: Purpose and tolerances, audit findings and the public benefit report.

The same loop runs at three speeds, with more people at each.
  1. Automated checksProofs and drift tests
  2. Human reviewC-Gate board, every release
  3. Independent auditAudit Council, major versions
  4. External standardThird-party certification
  5. Public reportBenefit report, every year
Five layers of assurance around the engine, from machine speed to public view.

What it guards against

The book names four ways a self-improving system can lose continuity: correspondence drift, goal discontinuity, knowledge erosion and human severance. All four are failures of continuity, not of capability. The engine is where all four guards meet.

Rupture (Chapter 2)Guard in the loopCRIS system
Correspondence driftThe system's model slowly separates from the world.↔ Correspondence proofEach change tested against realityCorrespondence ProverObservation Mesh
Goal discontinuityIts purpose is reinterpreted at each rewrite until it survives in name only.C Purpose testA successor must keep its purposeC-Gate
Knowledge erosionVerified knowledge is overwritten by unverified assumptions.Σ Promotion rulesOnly verified claims persistKnowledge Ledger
Human severanceThe improvement loop closes around itself and people become bystanders.Eᵢ Human observationPeople stay a source of realityCIXObservation Mesh
The four ways a self-improving system can lose continuity, and the CRIS systems designed to guard against each. Select a guard to see how it is designed to work.

Status and evidence

This status is set by the CRIS team and links to the evidence behind it. The engine moves from research to prototype, pilot and live only as each level's evidence is published. The CRIS roadmap places the CRL Engine prototype in its 2027 horizon, "Prove the loop".

Current status: Status to be confirmed

LevelEvidence requiredWho sets it
ResearchPublished concept page and draft specificationCEO
PrototypeWorkflow A running on a test model, with ledger recordsCEO, on the Labs and Systems leads’ advice
PilotPartner deployment on testnet and an Audit Council reviewCEO
LiveProduction use and an external audit passedCEO, with notice to the board

For developers: the Build path and the Continuity Invariant specification, draft 0.1.

Read more

Next steps

Owner: Head of Systems EngineeringNext review: 2027-03-01