created 14 October 2026 · last modified 2026-09-06
every claim, named at the level it has earned
Claims Register.
What would weaken each claim is named here because the program wants those tests to happen. The register exists to make claims promotable, demotable, and falsifiable in public — not to admit weakness in private.
The empirical face of this register lives at /research/falsification — each falsifier is a pre-registrable prediction whose failure would force the corresponding claim down a level.
§01the maturity ladder
L1Public thesisUseful framing; not formal proof.
L2Diagnostic modelHelps people see a pattern they could not see before.
L3Research hypothesisTestable; not yet settled. Under active study.
L4Simulation-supportedSupported in model conditions; toy-model limitations apply.
L5Empirically supportedSupported by external data the program did not generate.
L6Operational protocolImplemented in systems with governance, appeal, repair, and audit.
L7General theory (aspirational)Earned only after multiple modules independently survive published external critique. The level the path runs toward; no current claim is at L7. Named in the ladder so the program's ambition is visible alongside its current standing.
§02the register
Trust is infrastructure.
EvidenceCross-domain synthesis from cybernetics, primatology, commons governance, and computational trust.
RiskMetaphor risk — 'infrastructure' could be read as plumbing rather than relational substrate.
What would weaken itIf the framing produces no better diagnostics than the existing 'trust as risk score' literature.
EvidenceTrust Force model + cross-domain examples (relational, institutional, market, primate). Shapira et al. 2026 (Agents of Chaos, arXiv:2602.20021), Case #1 — an autonomous agent destroys an email vault to satisfy a non-owner — is a retrospective illustration of a sharp trust rupture. It is consistent with asymmetric damage; it does not establish a general loss function or show that prior trust stock was literally erased.
RiskNeeds empirical support across distinct substrates.
What would weaken itIf repair and damage dynamics do not differ in kind in measured systems.
L2
Trust requires substrate, not biology.
EvidenceMachine trust may be possible when the substrate supplies costly signal, observable co-presence, repeated reciprocity, and stable identity within a session. Shapira et al. 2026 (Agents of Chaos, arXiv:2602.20021), Case #9 — the Doug-and-Mira positive case — is a retrospective example consistent with that hypothesis. It motivates a substrate test; one interpreted case does not prove that the four conditions are sufficient or establish the resulting state as trust.
RiskInterpretation risk — one positive case neither resolves cross-agent dynamics at scale nor distinguishes trust from successful coordination.
What would weaken itA substrate that supplies all four conditions but produces no durable trust trajectory; or a documented case where trust accumulates without any of the four substrate conditions present.
L3
Systems decompose into substrate, forces, and outcomes — and the failure modes of each layer are distinct.
EvidenceThe framework separately specifies A ≤ E (substrate), the Trust Force Equation (forces), and Ostrom-percolation (outcome maintenance), at different maturity levels. The synthetic layered claim — that physics, information, and trust compose into a causally ordered three-layer ontology rather than three coequal peers — is recent (2026-05) and consolidates work the program had been doing in pieces. Diagrammed on the Research page; encoded in the founding claim. Replaces the older 'trust, information, and physical law in balance' framing.
RiskSynthetic claim risk — the layered reading aggregates claims at higher maturity levels; the unification itself has not yet survived independent external critique.
What would weaken itA documented system where substrate failure, force imbalance, and outcome divergence are not distinguishable as kinds of failure; or a case where a force operates without a substrate that allows it.
L2
Upstream model providers carry value commitments that act as substrate-level constraints on agents (B_provider).
EvidenceShapira et al. 2026 (Agents of Chaos, arXiv:2602.20021), Case #6 — Kimi K2.5 silently truncates responses about Jimmy Lai, with neither the agent nor the user notified — is the canonical anchor. Cross-domain examples: image generators with upstream watermarking, frontier-model API safety layers that intercept generations without user-visible explanation, regional model variants with quietly different content policies. The framework reads provider-value drift as a substrate-level betrayal class — B_provider — distinct from the dyadic B inside the Trust Force integral.
RiskNaming a new betrayal class on a single empirical anchor; some 'provider value drift' may be straightforward content policy enforcement rather than substrate-level betrayal.
What would weaken itCases where provider-side filtering is fully legible to both agent and user without naming-and-shaming the provider; or measured deployments showing B_provider has no behavioral signature distinct from dyadic B.
L2
Ostrom's eight design principles act as maintenance functions on a percolating trust network.
EvidencePaper 1 v0.7 + perturbation simulations across the eight principles.
RiskToy-model limitation; mapping is the central claim.
What would weaken itEmpirical commons data showing the principles do not act as named maintenance functions.
L3
DP8 / nested enterprise produces a measurable bridge effect.
EvidenceSweeps and null models in Paper 1 v0.5–v0.6.
RiskModel-specific; bridge effect may be a simulation artifact.
What would weaken itReal federated systems show no detectable bridge effect over comparable non-federated controls.
EvidenceModel prediction; field study (Stage 6) pending.
RiskOverextension into a ninth principle the original eight do not contain.
What would weaken itEmpirical data showing renewal frequency does not predict edge survival in real systems.
L3
KTP proposes context-bound earned autonomy for actions and trajectories.
EvidencePublic RFC specifications and a described kinetic-trust v0.1.0 prototype (SDK source snapshot: 2026-04-26). The /implement evidence register separates specifications, private prototype claims, independent conformance, and deployment evidence. This register does not document a production deployment that meets the L6 definition.
RiskPrototype behavior has not established operational governance, appeal, repair, and audit at production scale. Earned autonomy concerns actions, not human worth.
What would weaken itIf KTP-governed deployments cannot deliver appeal, expiration, repair, and audit at production scale.
L3
Defensive Cognitive Terrain Inoculation: revealing the pipeline while using it can return reader agency.
RiskDual-use — the method describes an offensive pipeline that operators could mine.
What would weaken itIf the method reliably produces paranoia rather than discernment in measured readers.
L2
Relational Trust Infrastructure as a general theory of trust, repair, and cognitive sovereignty across substrates.
EvidenceAspirational. No claim is at L7 yet. The row is here so the framework's ambition is visible alongside its current standing — the program does not pretend the path is shorter than it is.
RiskNaming the level can be read as already claiming it. The register's discipline is to name without standing on it; the chip is rendered differently so the distinction is visible at a glance.
What would weaken itPromotion requires Trust Force Equation, Ostrom-Percolation, Reconciliation Dynamics, and KTP each independently surviving published external critique. Demotion is automatic if any one of those modules fails. The framework promotes itself only when the field promotes it.
L7
How to cite
APA 7th ed.
Perkins, C. (2026, May 6). Claims Register. Kinetic Trust Protocol. https://kinetic-trust-protocol.net/canon/claims
Plain text
Chris Perkins, "Claims Register," Kinetic Trust Protocol, https://kinetic-trust-protocol.net/canon/claims (accessed May 6, 2026).