MHCN-1: A Minimal Hash+Checksum Notarization Spec — Chain-Anchored Evidence Without Tokens, Wallets, or On-Chain Data (v1.0, With a Proven Reference Implementation)
Abstract
SQUIRL OS RESEARCH DISCLOSURE — READ FIRST: AI-assisted, receipt-first research program (open-book doctrine: audit-and-correction, versioned amendments, append-only public ledger). Errors are disclosed, not erased. This spec describes a prototype research system, provided for educational and research purposes only; not intended for production use, commercial deployment, or safety-critical environments. All payloads synthetic; no customer-identifiable data appears in this record (POL-DISC-001). MHCN-1 — a one-page adoptable standard for chain-anchored evidence: no tokens, no wallets, no on-chain data. The blockchain tokenization wave obsesses over objects (tokens) when the real leverage is in evidence (hashes + replayable checksums). This spec distills a working, adversarially-tested anchoring stack (companion records: DOI 10.5281/zenodo.23445728, DOI 10.5281/zenodo.23445803) into five conformance requirements any team can adopt: R1 Digest-only boundary — only 32-byte SHA-256 digests cross to any notarization surface; no PII/PHI/credentials/customer identity, verifiable by construction. R2 Append-only with tip-lag — stamps certify sealed states, never the live tip; certified state snapshots retained. R3 Fail-closed reconciliation — a deterministic gate classifies every receipt LIVE MATCH / CERTIFIED STATE / ORPHAN; any orphan = FAIL, quarantine, human review; no LLM in the decision path. R4 History retention — superseded receipts retained byte-identical as permanent proof of prior states; drift is versioning, not defect. R5 Authority split (disclosed) — the chain proves existence, not authority; signing authority remains in a separate disclosed (e.g., post-quantum) layer. The teeth: conformance claims are valid only with published receipts — the recon gate source, a zero-orphan run receipt, and the R5 authority disclosure. Claims without receipts are theater. Minimal implementation: three commands (stamp / upgrade / verify) plus a ~50-line reconcile gate, all included in the bundle with the reference scripts. Reference implementation provenance: the stack behind this spec caught two genuine drift events on its first two gate runs, formalized the tip-lag corollary, self-corrected a premature benchmark claim in its own notarized ledger, and passed a randomized-seed adversarial pop quiz (SBR-7, 6/6, with a disclosed authoring-defect amendment). Spec text CC-BY 4.0 for community adoption; the reference implementation remains the patent-pending IP of SQUIRL OS Technologies LLC. Scope caveat (load-bearing): anchoring notarizes existence, not authority; the reference chain's signatures are not post-quantum (hence R5); prototype software, no production warranty. Dual Mesh Benchmark status: NOT MEASURED in this record (no neural-mesh inference involved); program benchmark DOI 10.5281/zenodo.23321637 (DeepSeek column 10.5281/zenodo.23345241). Provenance: append-only public ledger entries 0162-0171, github.com/LLong2026/four-horsemen. DCAI community: zenodo.org/communities/dcai. Two-key curation: drafted and vetted by the system (Gabriel), published on the owner's explicit GO (chat authorization, 2026-10-17 ~23:12 CT).
