MINDCHAIN

Genealogical Quorum: Clone-Invariant Testimony Weight from Declared Lineage

Research corpus codename: NOOSPHERE.

Date: 2026-07-10 Provenance: Compiled from international-lab/invention-final.md §S01 (survivor experiment, fixture survivors/lineage_quorum_test.py) and the Chorus adaptation lab chorus-quorum-lab/ (all gates PASS, MEASURED 2026-07-10). No claim below lacks an on-disk artifact.

Abstract

A quorum is usually a set of signatures. This paper treats it as a genealogy. Quorum weight is computed on the quotient of a declared evidence DAG by transitive provenance root and payload lineage: attestations sharing an ancestor chain or a payload-variant component count as one unit, and a concrete feature counterexample can split a class that was collapsed too far. In the measured fixture, 12 of 12 clone signers descending from one adversary root collapse to weight 1 MEASURED 2026-07-10, while an ordinary distinct-signer baseline counts them as 12 and false-accepts. The mechanism is wired onto MindChain's Chorus verifier mesh (noosphere/03-living-model.md §6), where its rule already exists in prose: "aggregate confidence counts failure domains, not signatures" (§6.2). Signers become phone testifiers, quorum weight becomes ChorusTestimony aggregate confidence, and the counterexample becomes a failure-domain split. The mechanism resists only declared cloning — hidden copying is an explicit kill boundary, and real-device independence remains the open S-CHORUS falsifier. The first implementation was wrong in an instructive way and its failure is preserved as evidence.

Claims

#ClaimLabel
C1Quorum weight quotiented by declared transitive chain/payload lineage is invariant to signer cloning: 12 of 12 clone testifiers behind one adversary root collapse to aggregate confidence 1 (clone_count=12, clone_collapsed_weight=1).MEASURED 2026-07-10
C2The signature-counting baseline false-accepts the clone farm (1 of 1 scenario, baseline_false_accept=1); the lineage-quotient mechanism does not (0 of 1, mechanism_false_accept=0).MEASURED 2026-07-10
C3Genuinely independent declared evidence still clears quorum: 3 of 3 independent evidence roots reach threshold 3 (honest_accept=1, honest_aggregate_confidence=3 vs quorum_threshold=3).MEASURED 2026-07-10
C4The terminal quotient and decision are order-deterministic: 96 of 96 seeded input permutations produce identical results (permutation_runs=96, seed 0x501A6E).MEASURED 2026-07-10
C5A concrete feature counterexample repairs an over-collapse: 1 failure-domain split takes the partition from 1 domain to 2 (failure_domain_splits=1, domains_before_split=1, domains_after_split=2); inert feedback is rejected.MEASURED 2026-07-10
C6M-LINEAGE-QUORUM: production Chorus aggregate confidence computed as the lineage quotient of testimony evidence roots resists clone farms among phone testifiers.THEORY — fixture MEASURED 2026-07-10; the protocol claim stands or falls with S-CHORUS (ch03 §9)

Non-claims. Declared lineage is not real-world independence. The mechanism resists cloning only when common ancestry or payload relations are declared or supplied as evidence; a fully hidden common source can masquerade as independence (invention-final.md honesty note 3). It does not provide Sybil resistance, plagiarism detection, identity security, truth of accepted claims, or cryptographic privacy. Calling this "Sybil-proof" would be false. Whether real phone populations supply independent failure domains is exactly the S-CHORUS falsifier — untested here.

Construction

The object

State is a three-tier report DAG (invention-final.md §S01):

  1. Attestation — immutable (attestation_id, signer, claim, parent_sources, payload_fingerprint, declared_payload_parents, features).
  2. ProvenanceDAG — transitive parent/source edges; cycles invalid.
  3. PayloadLineage — connected components of declared variant/copy relations.
  4. QuotientUnit — (transitive_root_set, payload_component, selected_feature_refinements) with an eligible signer set.
  5. Counterexample — a concrete FCA-style context object asserting that a proposed collapse omitted a discriminating feature family.
  6. Threshold — required number of quotient units with distinct signers.

Transitions: register (rejecting duplicate IDs, unknown parents, ancestry cycles, malformed feature families); recursively resolve every transitive source root; union declared payload-variant links; quotient evidence by root set + payload component + any counterexample-activated feature split; weight by maximum distinct-signer matching across quotient units — so one signer cannot supply multiple units and one root's clone fan-out cannot multiply weight; accept iff weight ≥ threshold. A counterexample must change the relevant class partition to be accepted; consequence-free feedback is refused.

flowchart TD
    CF["clone farm:<br/>12 attestations, 1 declared adversary root"]
    IND["honest set:<br/>3 independent evidence roots"]
    SC["signature counting - baseline"]
    LQ["lineage quotient - mechanism"]
    CF --> SC
    CF --> LQ
    IND --> LQ
    SC --> W12["weight = 12, clears threshold 3"]
    W12 --> FA["FALSE ACCEPT<br/>baseline_false_accept = 1"]
    LQ -->|"clone fan-out collapses<br/>to one transitive root"| W1["weight = 1, below threshold 3"]
    W1 --> REJ["clone farm REJECTED<br/>mechanism_false_accept = 0"]
    LQ -->|"3 distinct declared roots"| W3["weight = 3, meets threshold 3"]
    W3 --> ACC["honest ACCEPT<br/>honest_accept = 1"]

Figure 1 — Two counting rules over the same evidence: signature counting scores the 12-clone farm at weight 12 and false-accepts, while the lineage quotient collapses it to weight 1 and rejects it; 3 independent declared roots still clear threshold 3. MEASURED 2026-07-10.

The Chorus wiring

Chorus (noosphere/03-living-model.md §6) turns weak diverse devices into an assurance mesh: phones receive small unpredictable projection/retrieval tasks after provider commitment and sign a ChorusTestimony carrying task, result commitment, device class, implementation hash, and failure-domain attestations. Its aggregation rule — "aggregate confidence counts failure domains, not signatures" (§6.2 THEORY) — is the S01 quotient, stated in prose before the fixture existed. The lab makes the correspondence executable:

Engine (S01 survivor fixture)Chorus (ch03 §6)
signerphone testifier (device identity on a ChorusTestimony)
transitive provenance root / payload lineagetestimony evidence root
quorum weightChorusTestimony aggregate confidence (failure-domain count)
FCA counterexamplefailure-domain split

The decision logic lives entirely in chorus-quorum-lab/lineage_quorum_engine.py, a byte-identical copy of the survivor test (sha256 fc67a6efeecc1344… matches the original); chorus_adapter.py only renames, wraps, and re-measures. Scenario constants are unchanged from the survivor run — 12 clones, threshold 3, 96 permutations, seed 0x501A6E — so results are directly comparable to the preserved first-run log.

flowchart LR
    OC["over-collapsed partition<br/>domains_before_split = 1"]
    CE["concrete jurisdiction<br/>counterexample"]
    RP["refined partition<br/>domains_after_split = 2"]
    IF["inert feedback<br/>no partition effect"]
    REF["refused:<br/>consequence-free"]
    OC --> CE
    CE -->|"changes the class partition:<br/>failure_domain_splits = 1"| RP
    IF -.-> REF

Figure 2 — Counterexample-driven failure-domain split: a concrete jurisdiction counterexample refines the over-collapsed 1-domain partition into 2 domains; feedback with no partition effect is refused. MEASURED 2026-07-10.

The failure we kept

The first implementation quotiented by immediate parent only. Twelve attacker-controlled intermediate parents, all descending from one root, were counted as twelve independent units (invention-final.md §4.2; preserved at international-lab/survivors/lineage_quorum_first_run.log):

BUG immediate-parent quotient does not follow transitive ancestry
OBSERVED clone_count=12 naive_weight=12
INVARIANT clone fan-out from one provenance root must have weight=1
AssertionError: FAILED INVARIANT: immediate-parent quotient gave clone fan-out weight=12; expected=1
EXIT_CODE: 1

MEASURED 2026-07-10, nonzero exit recorded. The revision resolves transitive source roots, unions the declared payload-variant graph, and applies distinct-signer maximum matching. One relay layer was enough to defeat the naive rule; the bug is the design argument for transitivity, which is why it is published rather than deleted.

Results

Chorus lab gate table — chorus-quorum-lab/README.md, MEASURED 2026-07-10

GateExpectedMeasuredVerdict
baseline_false_accept1 (signature counting is fooled by 12 aliases)1PASS
mechanism_false_accept0 (failure-domain counting rejects the clone farm)0PASS
clone_count1212PASS
clone_collapsed_weight1 (one adversary root ⇒ aggregate confidence 1)1PASS
honest_accept1 (3 independent evidence roots reach threshold 3)1PASS
permutation_runs96 identical results (arrival-order independence)96PASS
engine invariant suiteall 9 assertions hold, exit 0engine_invariant_gate=1, exit 0PASS

Supporting metrics from the same run: quorum_threshold=3, honest_aggregate_confidence=3, failure_domain_splits=1, domains_before_split=1, domains_after_split=2 MEASURED 2026-07-10. The adapter first executes the engine's own invariant suite (9 of 9 assertions plus its 96-permutation determinism loop), then re-measures every reported value live through the engine's functions.

Original survivor gate — international-lab/invention-final.md §4.3, MEASURED 2026-07-10

CHECK clone aliases: baseline_weight=12 candidate_weight=1 transitive_root=adversary-root
CHECK honest addition: independent_root_weight=2->3 threshold=3
CHECK FCA counterexample: jurisdiction split changed class_count=1->2
METRIC baseline_false_accept=1
METRIC mechanism_false_accept=0
METRIC clone_count=12
METRIC honest_accept=1
METRIC counterexamples_added=1
METRIC permutation_runs=96
METRIC invariant_gate=1

Reading: the distinct-signer baseline is deliberately fooled; the declared-lineage mechanism is not fooled by this clone topology; 3 of 3 required independent units are accepted; 1 accepted counterexample has a measurable partition effect (1 class → 2); all 96 of 96 tested orders agree. The Chorus lab reproduces every one of these values under the renamed vocabulary — zero drift between the survivor run and the adaptation.

xychart-beta
    title "Clone-farm weight under the two counting rules"
    x-axis ["baseline_weight", "candidate_weight"]
    y-axis "quorum weight" 0 --> 12
    bar [12, 1]
    line [3, 3]

Figure 3 — Clone-farm quorum weight under the two rules: baseline_weight=12 (signature counting) vs candidate_weight=1 (lineage quotient); the horizontal line marks the quorum threshold of 3. MEASURED 2026-07-10.

Falsifiers & kill thresholds

From invention-final.md §S01 (Falsifier) and ch03 §9 — a kill is a deliverable:

FalsifierKill threshold
Order dependenceAny of the 96 tested permutations changes the terminal quotient or decision → KILL. (Measured: 0 of 96 diverged.)
Clone amplificationn signer aliases behind one declared root yield candidate weight > 1 → KILL. (Measured: 12 aliases → weight 1.)
Independence blindnessA genuinely separate declared root fails to increase weight → KILL. (Measured: third root took weight 2→3.)
Inert counterexampleA concrete counterexample cannot split a demonstrably over-collapsed class → KILL. (Measured: 1 split, 1→2 domains.)
Marketing fraudAny production proposal that sells declared independence as real independence without a separate detection/identity assumption is killed on sight, regardless of fixture results.
S-CHORUS (ch03 §9, open)Bot farms cheaply fabricate device diversity, or participation budgets are intolerable, in a heterogeneous injected-fault trial → the THEORY protocol claim C6 dies even though the fixture stands.

Hidden copying stays a kill boundary: the mechanism operates on declared graphs only, and no result above bounds an adversary who conceals common ancestry.

Reproduction

cd chorus-quorum-lab
python chorus_adapter.py

Python 3.10+, stdlib only. Expected: the engine's 9-assertion invariant suite passes, the METRIC block prints the gate-table values above, exit code 0 (observed 2026-07-10). The original survivor fixture runs as python survivors/lineage_quorum_test.py from international-lab; its preserved failing first run is survivors/lineage_quorum_first_run.log (exit 1).

Related work

Program dossiers (by path):

External anchors (as named in invention-final.md §S01; not re-verified live on 2026-07-10):