Claims ledger
Evidence
Every claim this site makes, with its label and what backs it. A tested claim names a test in the repository; the test suite fails if that test is missing. A label changes only when linked evidence exists.
tested 47proposed 10 · Download claims.v1.json
| ID | Claim | Label | Evidence |
|---|---|---|---|
| C-01 | The gate returns EXECUTE only when every check is PASS. | tested | tests/gate.test.ts“EXECUTE only when every check passes” |
| C-02 | Missing inputs produce UNMEASURABLE or UNAVAILABLE, never PASS. | tested | tests/gate.test.ts“missing inputs are never PASS” |
| C-03 | Each synthetic example produces its recorded outcome and reason codes. | tested | tests/gate.test.ts“examples reproduce expected determinations” |
| C-04 | A determination recomputes to the same digest and receipt (self-replay, not independent replay). | tested | tests/gate.test.ts“self-replay matches” |
| C-05 | A reassessment references the prior receipt and leaves the prior determination unchanged. | tested | tests/gate.test.ts“reassessment links prior receipt” |
| C-06 | Every archetype a class names has a registry entry and a detector. | tested | tests/appraisal.test.ts“every class archetype has a detector” |
| C-07 | The appraisal output contains no score, price, premium or rating. | tested | tests/appraisal.test.ts“appraisal has no score fields” |
| C-08 | Classes, archetypes and examples validate against their published JSON Schemas. | tested | tests/registry.test.ts“registry validates against schemas” |
| C-09 | Signed receipts verify, and fail verification after tampering. | tested | tests/signing.test.ts“signed receipt verifies and detects tampering” |
| C-10 | Portfolio concentration and the epistemic IBNR count are computed from determinations. | tested | tests/portfolio.test.ts“portfolio reports shared nodes and IBNR” |
| C-16 | The study answer key is generated from the gate's checks and matches the hash recorded in the preregistration. | tested | tests/study.test.ts“published answer key matches the gate” |
| C-17 | Threshold checks report headroom to their declared limit, measured from the worst end of any declared range. | tested | tests/headroom.test.ts“threshold checks report headroom from the worst end of the range” |
| C-18 | A declared range that straddles a limit is UNMEASURABLE, never PASS. | tested | tests/headroom.test.ts“a range straddling a limit is UNMEASURABLE, never PASS” |
| C-19 | Each threshold check states the value at which its result would flip, and that value does flip it. | tested | tests/headroom.test.ts“the flip threshold really flips the check” |
| C-20 | Every crossing class declares whether its consequence is bounded, and the conditions export carries it. | tested | tests/headroom.test.ts“every class declares its consequence bound, and exports carry it” |
| C-22 | When several passing checks tie for least headroom, the appraisal lists all of them rather than picking one. | tested | tests/headroom.test.ts“appraisal lists every check tied at the nearest limit” |
| C-23 | The receipt chain verifies, and detects an edited, removed, reordered or mismatched record when checked against the original records or head hash. | tested | tests/recorder.test.ts“chain verifies and detects tampering” |
| C-27 | A chain rewritten with recomputed hashes still verifies on its own; only a head hash kept elsewhere detects it. | tested | tests/recorder.test.ts“a recomputed chain is caught only by a head hash kept elsewhere” |
| C-24 | The insurer view carries no names, sources or proposal text from any example proposal. | tested | tests/recorder.test.ts“insurer view carries no identifying text from the proposal” |
| C-25 | An irreversible crossing holds unless its class fails closed, and a fail-silent class never passes. | tested | tests/recorder.test.ts“irreversible crossings require a fail-closed class” |
| C-26 | A crossing executed before its determination counts as epistemic IBNR even when the determination is EXECUTE. | tested | tests/recorder.test.ts“execution before the determination counts as epistemic IBNR even when EXECUTE” |
| C-28 | Evidence below a class's minimum tier fails, and evidence without a tier is unmeasurable. | tested | tests/source-spec.test.ts“evidence below the class's minimum tier fails, and a missing tier is unmeasurable” |
| C-29 | A class that breaks the source's I2 admissibility rules cannot warrant any crossing. | tested | tests/source-spec.test.ts“classes that break the source's I2 rules cannot warrant any crossing” |
| C-30 | A determination uses only evidence available at its record time, so re-running at an earlier time reconstructs what was known then. | tested | tests/source-spec.test.ts“a determination uses only evidence available at its record time” |
| C-31 | Late evidence can defeat a crossing's standing while the original determination stays unchanged and still replays. | tested | tests/source-spec.test.ts“late evidence defeats the original standing without rewriting it” |
| C-32 | Standing changes are labelled PRESERVED, STRENGTHENED, WEAKENED, DEFEATED or NEW with their causes. | tested | tests/source-spec.test.ts“standing effects cover revision, weakening and new determinations” |
| C-33 | Insurer exports coarsen timestamps to the hour and suppress any class with fewer than k crossings. | tested | tests/source-spec.test.ts“insurer exports coarsen time and suppress classes below k” |
| C-34 | The evidence tiers, class dimensions and I2 rules match crossing-graph-spec v0.4 as the source intends them for real domains. | proposed | none yet |
| C-35 | The cover scenario validates against its published schema. | tested | tests/cover.test.ts“cover scenario validates against its schema” |
| C-36 | The cover trigger fires only when reliance, failure before T, a defined loss and a declared causal link are all present; any one missing leaves it NOT_TRIGGERED. | tested | tests/cover.test.ts“an insured event needs reliance, failure before T and a defined, attributable loss” |
| C-37 | The joined record reports, per stage, how many crossings have observed data; reliance, consequence and loss are observed for one of twelve examples. | tested | tests/cover.test.ts“completeness shows the later joins are mostly missing” |
| C-38 | Every applicable mutant of each class's warranted example holds for exactly its expected reason. | tested | tests/cover.test.ts“every applicable mutant holds for exactly its expected reason” |
| C-39 | A joined dataset from configuration through reliance, consequence and loss, collected in real deployments, can support pricing. | proposed | none yet |
| C-40 | Value status and structure status move independently: a pending amount raises the first, an unpathed independent observation raises the second. | tested | tests/uncertainty.test.ts“value and structure move independently over the INS-005 timeline” |
| C-41 | Structure is untested by default and supported only after a passed structural test; a failed test or an unpathed consequence challenges it. | tested | tests/uncertainty.test.ts“an absence of challenges is untested, never supported” |
| C-42 | Hidden coupling challenges structure without changing value status. | tested | tests/uncertainty.test.ts“hidden coupling challenges structure without touching value status” |
| C-43 | The uncertainty pair contains no number and is never collapsed into a single score. | tested | tests/uncertainty.test.ts“the pair is never collapsed into a single number” |
| C-44 | A triggered cover result carries an explicit structural caveat. | tested | tests/uncertainty.test.ts“a triggered cover result carries the structural caveat” |
| C-45 | The declared consequence paths of each class contain the consequences that matter in real deployments. | proposed | none yet |
| C-46 | A published class passes syntactic admission (schema, I2, battery); admission never claims domain validation. | tested | tests/workspace.test.ts“admission is syntactic: every published class passes schema, I2 and its battery” |
| C-47 | Without a warranted example the battery is reported unavailable, not passed. | tested | tests/workspace.test.ts“without a warranted example the battery is unavailable, not passed” |
| C-48 | Schema errors and I2 violations are refused, and refusals stay on the workspace record. | tested | tests/workspace.test.ts“schema errors and I2 violations are refused, and refusals stay on the record” |
| C-49 | A workspace file serialises to stable bytes and every determination replays from the file alone. | tested | tests/workspace.test.ts“serialises to stable bytes and replays from the file alone” |
| C-50 | Replay fails visibly when class or policy material is missing, or a proposal or chain record is edited. | tested | tests/workspace.test.ts“replay needs the class and policy bytes, and any edit is visible” |
| C-51 | An evidence pack is byte-for-byte reproducible from its workspace and verifies against its manifest. | tested | tests/workspace.test.ts“is byte-for-byte reproducible from the workspace and verifies” |
| C-52 | Pack verification reports tampered, missing and extra files and a changed manifest. | tested | tests/workspace.test.ts“reports tampered, missing and extra files, and a changed manifest” |
| C-53 | No application source calls a network API. | tested | tests/workspace.test.ts“no application source calls a network API” |
| C-54 | Each science claim reports its own value and structure status; an unexecuted path challenges structure. | tested | tests/science.test.ts“each claim reports its own value and structure status” |
| C-55 | Disagreeing mappings end UNRESOLVED, and a later convention does not rewrite that outcome. | tested | tests/science.test.ts“disagreeing mappings end UNRESOLVED and a later convention does not rewrite that” |
| C-56 | No application source calls a network API, and the exported site loads no third-party resource; a browser run of the workflow recorded no requests (docs/LOCAL_FIRST_CHECK.md). | tested | tests/workspace.test.ts“no application source calls a network API” |
| C-57 | Pilots using the workspace are a viable paid offer. | proposed | none yet |
| C-21 | The declared consequence bounds are correct for real instances of each class. | proposed | none yet |
| C-11 | The ten risk archetypes capture the main forms of epistemic risk in AI-mediated crossings. | proposed | none yet |
| C-12 | Claims professionals reconstruct decisions faster and more accurately from receipts than from ordinary logs. | proposed | public/study/PROTOCOL.md |
| C-13 | The crossing is a usable exposure base for AI-mediated decisions. | proposed | none yet |
| C-14 | Epistemic IBNR is a leading indicator of AI-related loss. | proposed | none yet |
| C-15 | The crossing classes reflect real insurer, accounting or model-risk practice. | proposed | none yet |
What "tested" covers here
The tested claims are about the machinery: the gate, appraisal, portfolio, signatures and schemas behave as described on synthetic inputs. None of them says anything about real-world risk. The upstream conformance fixture and its verifiers are documented on Warranted Crossings.
To re-run the tests from a checkout: npm test