CAIN-42 · Authority Notary v3
CAIN-42 is a Byzantine trust fabric. The Authority Notary is its independent verification surface: six checkable layers that answer who authorized this, may it proceed, is the agent's world still true, did the effect match the authorization, and can a stranger prove it without trusting us. A quorum co-signing is never an authorization.
Each layer can only say no. The stack is AUTHORIZED only when every layer is, and a missing layer is named rather than assumed to pass.
loading the live layer descriptors…
These buttons call the real routes on this host. Both outcomes are shown on purpose: a notary that only ever says yes is a rubber stamp.
not run yet
not run yet
admin:*.
not run yet
not run yet
The bundle ships its own verifier. It imports no CAIN code and re-derives every verdict
instead of reading our ok flags.
pip install cryptography curl -O https://cainstudio.online/notary/v3/evidence/bundle curl -O https://cainstudio.online/notary/v3/evidence/verifier.py python3 verifier.py bundle --pinned-pub <the key you pinned>
Every subsystem in the published architecture, graded from the tree rather than from a document. A row is REAL only when production code exists and a test exercises it. This is the page that makes the marketing copy auditable.
loading the matrix…
| Claim | Status | Basis |
|---|---|---|
| L2 decision algebra is correct and fail-closed | PROVED (bounded) | Z3, five theorems, UNSAT of every counterexample |
| L2 evaluator ↔ independent Lean semantics | PENDING | Lean 4 is not installed on the build host; not claimed |
| L1 chain only narrows, offline | TESTED | PEDIGREE/APS/AIP/WIMSE invariants, adversarial tests |
| L3 state is tamper-evident and divergence is localised | TESTED | per-tenant SHA-256 chain, rewrite and gap tests |
| L4 effect stays inside the mandate | TESTED | widening, budget, replay, expiry and settlement tests |
| L0 quorum is equivocation-resistant | TESTED | N=4, Q=3, Sybil and equivocation tests |
| Third-party audit / SOC 2 / customer run | NOT MET | same-operator evidence only |