Research · knowledge architecture
Shared Object Networking
Persistent knowledge for AI, with context, provenance, and room to evolve.
The idea
Shared Object Networking (SON) explores a knowledge architecture that separates reusable facts and objects from the contextual relationships and inferences built around them. The goal is to let AI systems work with persistent, connected knowledge that can be updated without retraining the underlying model.
The papers below trace the idea from an object-based knowledge model to a distributed GraphRAG architecture, with attention to semantic alignment, provenance, trust, and keeping private knowledge under local control.
Read the papers
Both papers are hosted here as original PDFs. Start with the whitepaper for the underlying model, then the GraphRAG paper for the distributed architecture.
Shared Object Networking
A Model for Knowledge Representation for use in AI Systems
Introduces modular knowledge objects and layered relationships, separating declarative facts from contextual and inference layers. It explores persistent knowledge, incremental updates, and retrieval-augmented generation.
Read the whitepaper (PDF) Download whitepaper
Date shown on the cover: February 22, 2025. The original source filename is dated February 20, 2025.
Distributed GraphRAG via Shared Object Networking (SON)
Develops the model into an architecture for retrieval and reasoning across heterogeneous knowledge graphs. Topics include hybrid z-axis mapping, agentic knowledge discovery, evidence and confidence tracking, privacy boundaries, and evaluation criteria and use cases.
Continuing the research
These papers document the development of my thinking about SON and distributed knowledge systems. They provide a foundation for further implementation, evaluation, and discussion.