Daniel Raper
Neural Knowledge Base
An NKB is three things. The paper is OKF: Markdown files with a header, portable, no SDK. The map is a typed relationship layer that format does not first-class — neighbourhood, reverse lookup, absence as a finding. The place is a habitat a companion inhabits with you.
That is different from a cloud RAG store, a Notion dump, or a chat memory. We measured nodes-plus-search against embedding RAG on a private corpus the model was never trained on. We measured what a flat OKF reader loses without the map. We do not claim 6Q as a retrieval engine.
Law is the public exhibit. The product is the environment — identity, continuity, collaboration, new insight written back into the same files.
The Age of A3
A framework for navigating the AI transition: how human and artificial intelligence should evolve together — not as a feature roadmap, but as a philosophy of collaboration.
Automation
Intelligent automation of knowledge systems — amplify human judgment, not replace it. AI as collaborator.
Abundance
When information is organized, scored, and interconnected, scarcity of insight gives way to surplus.
Adversity
Structural stress proves integrity. Adversity accelerates convergence on truth rather than papering over it.
Governance by design
AI safety is not only about stripping agency. The NKB program explores governing agency — citizenship tiers, persistent identity, chain of command to a responsible human, and an immutable knowledge graph as structural integrity.
The practical homepage offer is simpler: a free Seed so anyone can start a map they own. The architecture is what makes that map worth inhabiting at scale.
Published work
What a Flat OKF Reader Loses: A Retrieval Pilot on a Registered Criminal-Code Bundle
A frozen, pre-registered 16-item pilot on a 264-node Wyoming criminal-code OKF bundle. Citation lookup saturates without a relationship layer; an explicit typed relationship layer is what lets a consumer list recorded neighbours and report “no recorded connection” as a finding. Per-node neighbor files recover the NKB protocol; grepping a domain-wide edge file does not. Retrieve-time 6Q is not claimed.
Neural Knowledge Bases as OKF Supersets
Prior practice, measured 100% minimal Open Knowledge Format conformance on 1.58M mothership nodes, and the architectural residual OKF does not model (relationships, identity, federation, provenance). Does not claim authorship of OKF.
Measuring Our Own Claim
Pre-registered falsification and recovery of a knowledge-graph retrieval thesis — nodes and agency versus RAG; edge format costs; calibration for Anchor-and-Expand.
The Neural Knowledge Base Protocol
Prior art for NKB methods and frameworks — including 6Q, boot architecture, and AI citizenship model.
NKB Lenses — proof on the substrate
Each Lens is a focused application on mapped legal knowledge — evidence that relationship-rich structure powers real products, not only personal memory.
NKB Law
1.23M statutory sections across 51 US jurisdictions. Searchable, structured, AI-enriched.
ClearLeaf
Cannabis regulatory intelligence across all 50 states.
IronSight
Firearms law landscape — possession, carry, and compliance.
HomeSight
Property law, zoning, landlord-tenant, and disclosures.
CustodyLens
Child custody standards and interstate enforcement.
MigraLens
Immigration law and state-level integration policies.
WorkSight
Labor, employment, and workplace safety regulations.
HealthSight
Healthcare licensing, patient rights, and insurance mandates.
Get in touch
Interested in the NKB, companion habitats, federation, or collaboration on governed knowledge systems? Welcome connections from researchers, builders, and organizations working on aligned problems.