Holistic Organizational Intelligence Platform with Deterministic Query System
The architecture behind Ripplemesh — a unified experience ledger with semantic context embedded at the moment of writing, privacy-segregated identity, multi-class actors, and a natural-language query system that is structurally incapable of fabricating a fact — is now the subject of a U.S. non-provisional patent application.
Application No.
19/796,012
Filed
September 3, 2026
Type
Utility — Non-Provisional, 35 U.S.C. § 111(a)
Priority
U.S. Provisional 63/904,593 · Oct 24, 2025
Claims
40 total · 9 independent
Inventor
Randy Stewart Miller
Assignee
Ripplemesh Corporation
Abstract
Organizational activity across learning, operations, maintenance, safety, and management is persisted as experience statements in a unified ledger. On each entity-change event, deterministic rules keyed on a governed verb identifier and object type resolve the event to concepts in a governed ontology and embed the resulting semantic context, with per-edge confidences, into the statement at persistence time, so cross-domain grouping needs no post-hoc enrichment and embedded context is immutable against later rule changes. Actors may be human participants, software agents, automated processes, robotic platforms, or biological assets, all resolved by one audited function. Statements carry only non-identifying identifiers; personal information resides in a separate schema reachable solely through an audited, role-verified function. Natural-language access uses a classification stage emitting no factual assertion, a margin-gated ambiguity check, and a deterministic retrieval stage that returns a data-missing indication rather than a fabrication.
Nine Independent Claims
What the application protects
Semantic Enrichment at Persistence
Every experience statement is tagged to a governed ontology by deterministic, auditable rules the instant it is written — no post-hoc classification pass, and historical context is immutable when rules change.
Two-Part Deterministic Query
A language model classifies intent but is structurally barred from stating facts. A deterministic retrieval stage — with no language model in it — produces every factual answer, with provenance.
Ambiguity-Aware Routing
Requests are matched against every candidate, scored, and margin-gated. Under-determined questions return a clarification instead of a confident answer to a question that was never asked.
Multi-Class Actor Attribution
Humans, AI agents, automated processes, robots, plants, and animals all record activity into one ledger and are resolved by a single audited function — enabling cross-class reasoning in one query.
Privacy-Segregated Identity
The analytical ledger contains no personally identifying information. An audited, role-verified resolution function is the only path from a stable identifier to a person.
Competency Decay Scoring
Credential age, peer-endorsement recency, and execution recency are combined into a predictive decay score — emitted with the statement identifiers that prove it.
Cross-Domain Readiness Traversal
Worker → credential → qualification → procedure → active permit, with confidence carried by the evidence for each link rather than by the credential record itself.
Procedure Drift Detection
Recorded executions are compared to the authored step sequence. Consistent deviation across many operators flags the procedure — not the people — as defective.
Tacit-Knowledge Capture
Structured expert interviews are persisted as concept nodes and edges in the same governed ontology, so expertise remains queryable after the expert retires.
Filed Document
Non-Provisional Application & USPTO Receipt
39 pages — acknowledgement receipt, specification, 40 claims, abstract, and 11 figures.
"Patent Pending" indicates that a patent application has been filed with the United States Patent and Trademark Office and is awaiting examination. No patent has yet been granted. Subject matter first disclosed in the provisional application is entitled to the October 24, 2025 filing date.