Observe Only
NYRV observes and records. It never decides, never acts, never judges. Recognition without authority.
Neural Yielding Recognition Vector
Independent provenance infrastructure for automated decisions. Producers commit to each decision; NYRV records a commitment it cannot read; and anyone can confirm, in open-source code, that a record binds to exactly the decision claimed — with no trust in NYRV required.
Every compliance platform stores its evidence in its own database. That's not proof — that's a promise. NYRV is the independent layer underneath: content-blind, and built so that what a record binds to doesn't rest on NYRV's word. You check the math yourself.
The Problem
Automated systems are taking consequential actions. Hiring decisions. Credit determinations. Insurance underwriting. Content moderation.
When regulators, auditors, courts, or counterparties demand to see what an automated system did — and proof the record can be trusted — organizations hit a structural gap. Today that evidence lives in vendor databases, self-attested logs, and governance dashboards. None of it is independently verifiable.
That gap is becoming a liability:
The open question was never whether you must keep records. It's whether anyone can trust the records you kept.
Who It's For
Organizations subject to NYC LL144 bias-audit requirements and EU AI Act Article 12.
Lenders facing CFPB scrutiny under ECOA, FCRA, and fair-lending statutes.
Carriers using AI in underwriting, pricing, and claims subject to state regulator examination.
Platforms operating under the EU Digital Services Act and emerging transparency requirements.
Vendors whose customers increasingly require third-party-verifiable evidence beneath the platform.
Auditors who need evidence they can independently verify, not dashboards they have to trust.
Core Directives
These are not policies — they are constraints in code, checked by tests.
NYRV observes and records. It never decides, never acts, never judges. Recognition without authority.
Events are added, never modified or deleted — an append-only schema enforced in code and in the database. No update path. No delete path.
NYRV records a content-blind commitment and producer-chosen labels — never the decision content, never field names. It can confirm a record exists. It never sees what it contained.
NYRV holds no financial or operational relationship with any party whose conduct it records. Today that independence is structural. Cryptographic independence — proof that even NYRV cannot rewrite the record — is the anchor layer, in design.
NYRV has hard limits. It cannot make autonomous decisions. It cannot execute actions. It cannot expand its own authority. Decision authority remains with the operators it serves.
The math is independent: anyone can run the open-source verifier, with no NYRV dependency, and confirm a record binds to the decision disclosed. Trust no one — check it yourself.
Architecture
What is tested is the architecture; implementation follows it, not the other way around. In force and in design are marked plainly.
Each record links to its predecessor by hash. Anyone can recompute the chain and confirm it is internally consistent.
PostgreSQL-backed, append-only schema enforced in code and at the database level. No update path. No delete path.
Each observation returns a receipt in the same call — record id, timestamp, and chain hashes. Your system confirms the record landed before it proceeds.
The producer commits client-side; the codebase has no path for decision content to reach the ledger. Privacy and a smaller trust surface, by construction.
Today a producer cannot prove NYRV did not alter a record, and the timestamp is NYRV-asserted. The layer that closes this — independent anchoring, so a record's integrity and timing do not rest on NYRV — is in design; the mechanism is a reserved decision. NYRV describes the property it is building toward, never claims it as built.
Cryptographic boundaries between organizations. Your records are yours alone. No commingling, by design.
Standards Alignment
NYRV operates VeritasChain Protocol (VCP) v1.1-conformant infrastructure. VCP is published by the VeritasChain Standards Organization (VSO), an independent, vendor-neutral standards body — an open, public, reviewable standard rather than a proprietary architecture.
That discipline means customers do not have to take NYRV's word for the architecture's correctness; the specification is public, licensed CC BY 4.0, and open to independent technical scrutiny. NYRV's value is the operated infrastructure that implements it.
The direction is industry-wide: the IETF's SCITT working group is standardizing exactly this class of transparency infrastructure.
NYRV's VCP v1.1 conformance commitment is documented in ADR-0002 in the repository. The operator-independence extension above the protocol is a reserved decision — not yet operated.
NYRV is one implementation; the protocol exists independent of any single operator.
Integration
Commit client-side. Send only the commitment. Verify offline. The flow is simple because the principle is.
Your system commits to each decision client-side with the open-source SDK — a salted hash — and keeps the disclosure record. The decision content never leaves your system.
NYRV receives only the commitment and your labels, chains it, and persists it to the append-only ledger. NYRV never sees the content.
NYRV returns a synchronous receipt — record id, timestamp, chain hashes — so your system confirms the record landed.
Later, anyone you hand a bundle to runs the open-source verifier and confirms — offline, with no trust in NYRV — that a disclosed decision binds to exactly that record.
Receipt Structure
Each accepted observation returns a receipt in the same call. The disclosure record you keep client-side is what later opens the commitment for verification.
{
"accepted": true,
"durable": true,
"rid": "<record-id>",
"timestamp": "<ISO-8601 UTC>",
"previous_record_hash": "<hash — present when durably chained>",
"record_hash": "<hash — present when durably chained>"
}The chain hashes are included once the record is durably chained. The commitment is never returned — it stays in your client-side disclosure record.
Current Status
The content-blind commitment ingest, the SDK, and the open-source verifier are in the repository. The operator-independence (anchor) layer and a public, self-serve verification path are in design. NYRV is not "production complete," and won't say it is.
Engaging design partners now. Organizations under active regulatory pressure with current evidence gaps are invited to take part. Reach out through the contact information below.
The codebase is open for inspection at github.com/Saabe-CEO/nyrv.
The code, the principles, the test invariants — all open for inspection. Read it. Run the tests. Verify what works.
github.com/Saabe-CEO/nyrv