Operator · Accumulation

Knowledge Ledger

A permanent record of what was learned, including what failed.

Approved by the Board for implementation · 20 November 2026

In one paragraph

The Knowledge Ledger is the permanent memory of every CRIS system. Every test result, every decision, every verified contribution and every rejected change is written to it, and nothing written is ever overwritten. Each new version of a system inherits the whole ledger, so it starts from everything its predecessors learned, including their mistakes. This is how we guard against knowledge erosion: the slow loss of what a system once knew.

It carries Σ, accumulation, the fourth operator of the Fourth Law. For every result it answers one question: has this earned a permanent place in the record?

Where it sits

The ledger carries the fourth stage of the loop, Accumulate. In the CRL Engine's Workflow A it records every result at step 4, whether the change passed or failed.

Knowledge today.
A more continuous tomorrow.
  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 →

  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.

Rules for this system are proposed in the AI Constitution, Article 11: lineage, rollback and archival. They are approved by the Board for implementation.

What gets written

Every system in the loop writes to the ledger, and nothing reaches the next cycle without an entry.

Knowledge Ledger ΣAppend-only. Read by the Reasoning Core, the Value Meter, auditors and the public benefit report.
Every system in the loop writes to the ledger; nothing reaches the next cycle without an entry.
EntryWritten byWhen
Correspondence proofCorrespondence ProverEvery candidate change, pass or fail
C-Gate decisionC-GatePromote, hold or reject
Rejected upgradeCRL EngineWith the contradictions that failed it
Verified contributionCIX, after the ProverWith its provenance and consent record
RollbackCRL EngineWhich version, why, and the result
IncidentSystems EngineeringLinked to every affected version

Contributions from people enter the same way. In Workflow B, a contribution to CIX reaches the ledger only after the Prover has verified it, and only then can the Participation Value Meter reward it.

  1. 1Join CIXVia the web or the CES app
  2. 2Consent and IDPrivacy and ownership
  3. 3ContributeAn observation or insight
  4. 4Verified?Correspondence Prover · ↔No → feedback, try again
  5. 5Enter ledgerKnowledge Ledger · Σ
  6. 6Value MeterScores the contribution
  7. 7RewardPaid; reputation rises
  8. 8Benefit reportAggregate and public
On CIX, only verified contributions earn rewards. More on CIX →

How it works

Every entry follows the same four steps.

  1. 1WriteAn append-only entry with lineage, evidence references and signatures
  2. 2PromoteBy rule: hypothesis → supported claim → verified knowledge
  3. 3ConsolidateCompress periodically without deleting history
  4. 4ExposeRead-only views to auditors; summaries to the public
TriggerAny proof, decision or verified contribution
OutputLedger entries each successor inherits
NeverNo entry is overwritten or deleted; corrections are new entries
The Knowledge Ledger workflow (CRIS document, Section 9).
  1. Write an append-only entry with its lineage, evidence references and signatures.
  2. Promote it by rule: from hypothesis, to supported claim, to verified knowledge.
  3. Consolidate periodically without deleting history (book Chapter 9.6).
  4. Expose read-only views to auditors and, in summary form, to the public.

The trigger is any proof, decision or verified contribution. Auditors can read the full ledger at any time; the public sees it in summary form.

Anatomy of an entry

Each entry says what it is, what it builds on, the evidence behind it and who wrote it. It also carries a link to the entry before it, so the record can be checked from end to end. The fields below are a draft; the entry format will be published as a specification.

Illustrative — draft fields, not a real entry
Ledger entry · draft fields
Entry ID
Unique and permanent
Type
Proof, decision, contribution, rejection, rollback or incident
Lineage
The version and earlier entries it builds on
Evidence references
Pointers to evidence, not personal data
Status
Hypothesis, supported claim or verified knowledge
Signatures
The system or person that wrote it
Time
When it was written
Link to previous entry
So any change to history is detectable
entry      Σ-000412
type       correspondence proof
parent     Σ-000398
lineage    version 3 → candidate 4
evidence   vault set 17 (reference)
status     supported claim
signed     Correspondence Prover
previous   Σ-000411 · fingerprint
What a ledger entry records. Field names are shown in plain words; the entry format will be published as a specification.

Three levels of knowledge

Not everything in the ledger is treated as fact. A new claim starts as a hypothesis. When it passes the Correspondence Prover it becomes a supported claim, and when it is repeatedly confirmed it becomes verified knowledge, the only ground our systems are allowed to reason from. The rules for each step are published.

  1. 1HypothesisAn idea or claim on record; not used as ground for reasoning
  2. 2Supported claimBacked by evidence that passed the Correspondence Prover
  3. 3Verified knowledgeRepeatedly confirmed; what the Reasoning Core may build on
A later contradictionis written as a new entry that points back; the earlier entry is never edited, and its level can be reviewed.
Promotion follows published rules. The exact thresholds for each step are set by Continuity Labs and will be published in the specification.
LevelWhat it meansWho uses it
HypothesisA claim on record, not yet testedResearchers; never the ground for reasoning
Supported claimBacked by evidence that passed the ProverThe Reasoning Core, with caution
Verified knowledgeRepeatedly confirmedThe Reasoning Core, as verified ground

The exact thresholds for each step are set by Continuity Labs and will be published in the specification.

Append-only and consolidation

Append-only means new entries are added and old ones are never changed. If later evidence shows an entry was wrong, we add the correction as a new entry that points back to it, so both remain visible. Each entry also links to the one before it, so any change to history would be detectable.

Illustrative — not real entries

The ledger

  1. Σ-101Claim recorded
  2. Σ-102Proof passes
  3. Σ-117New evidence contradicts it
  4. Σ-118Correction, pointing back to Σ-102

Never

  1. Σ-102Silently edited to match new evidence
  2. History lostNo one can see what changed
A mistake and its correction both stay visible. Each entry carries a link to the one before it, so any change to history would be detectable.

From time to time the ledger is summarised into a compact view for fast reading. This is consolidation, and it never deletes anything: the full history stays, and every summary records which entries it came from.

Consolidated viewWhat is currently known, in compact form, for fast reading
  1. Σ-101
  2. Σ-102
  3. Σ-103
  4. Σ-104
  5. Σ-105
  6. Σ-106
  7. Σ-107
  8. Σ-108
  9. Σ-109
  10. Σ-110
  11. Σ-111
  12. Σ-112
A consolidated view summarises entries without removing them, and records which entries it came from, so auditors can always go back to the source (book Chapter 9.6).

Failures are knowledge

Our rule is direct: rejected upgrades are written to the ledger, not deleted (book Chapter 8.6). When a change fails its tests, the failure is recorded with what was tried, the evidence that stopped it and why the gate said no. The next cycle reads that record before proposing anything new. A rejected change is not wasted: it is how a system remembers what not to do.

  1. Version 1Writes entries
  2. Version 2Inherits all of them, adds more
  3. Version 3Inherits all, including failures
  4. RollbackAny version can be traced and restored

Without a ledger

Each new version can forget what earlier ones learned: knowledge erosion.

With the ledger

Nothing verified is lost, and no failure is repeated unknowingly.
Every successor inherits the ledger: the guard against knowledge erosion, one of the four ruptures (book Chapter 2).

Every successor inherits the whole ledger, failures included, and any earlier version can be traced and restored. That is the guard against knowledge erosion, one of the four ruptures.

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.

Read about knowledge erosion on the ruptures page.

Who can read it

Our systems read the ledger to decide what to do next. Independent auditors can read all of it. The public sees summaries, including the figures in our public benefit report. No one, inside or outside CRIS, can edit or delete an entry.

Who can read the ledger, and how much. A view for participants of their own contributions is proposed and will be shown here only once it is approved.
ReaderWhat they seeBasis
Reasoning CoreVerified knowledge and open contradictionsWorkflow, Section 9
Participation Value MeterVerified contributionsWorkflow B
Auditors and the Audit CouncilThe full ledger, read-onlyStep 4
The publicSummaries and the benefit scorecardStep 4
No oneEditing or deleting an entryThe append-only rule

Auditors read the ledger at every one of the five layers of assurance, from automated checks to the public report.

  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.
Awaiting counsel. The paragraph on personal data and withdrawal of consent will appear here, in the wording approved by our privacy lead and legal counsel, before CIX sign-up opens.

Status and read more

This status is set by the CRIS team and links to the evidence behind it. The CRIS roadmap places the first CRL prototype, and with it the ledger's first entries, in 2027.

Current status: Status to be confirmed

LevelEvidence requiredWho sets it
ResearchConcept page and draft entry format publishedCEO
PrototypeEntries written by a test CRL cycle; append-only and links between entries testedCEO, on the Labs and Systems leads’ advice
PilotUsed in a partner deployment on testnet; auditors read it; summary views publishedCEO
LiveProduction use; external assessment passed; feeds the benefit reportCEO, with notice to the board

For developers: the ledger entry format, draft 0.1, and the Verify path.

The ledger is also where our public benefit indicators come from.

Knowledge LedgerRead-only summary views
  • Share of promotions with a complete proof, gate record and rollback testContinuity
  • Rejected upgrades and contradictions resolvedContinuity
  • Audit findings raised and closedGovernance
  • Verified contributions enteredPeople
Public benefit reportPublished every year
Indicators the public benefit report can read straight from the ledger. No figures are shown until they appear in the report.

Next steps

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