GraphNews

Ontologies are powerful and in the 2000’s they were a cutting-edge idea, but they were challenging to explain — most informaticists…

420+ people. 1,700+ public ties. Search anyone in the AI industry.

A JSON Schema is not an ontology. Everyone building LLM extraction pipelines is converging towards a JSON Schema (YAML same thing). It works really well, because function calling (say, Pydantic AI) and structured outputs all speaks JSON natively and you get typed, validated data out of your model. But there's a quiet assumption baked into that choice. A JSON Schema describes the shape of data, it has no concept of what the data means. Here are 4 things it simply cannot do and why they matter once your knowledge graph grows past a few hundred nodes. 1. Two extractions, two valid JSON objects, same real-world person. JSON Schema sees them as unrelated. OWL's functional properties and `owl:sameAs` let a reasoner merge them automatically. Without this, your graph accumulates ghost duplicates indefinitely. 2. A JSON Schema composes structure but it carries no semantic inheritance. `OWL's rdfs:subClassOf` means a reasoner knows a Researcher is a Person without you re-stating it everywhere. Add property chains and you get implicit relationships inferred from existing ones for free. 3. SHACL validates relationships between nodes — every Contract must have exactly one signatory who is a verified Person. That's a constraint you cannot express in JSON Schema at all. 4. Not the least...the open-world assumption. A JSON Schema treats missing information as an error. OWL treats it as unknown. In a knowledge graph built from thousands of documents, most entities are partially described. That's not a bug, rather the nature of real-world knowledge. The closed-world assumption is correct for API contracts. It's wrong for knowledge graphs. Now, of course I use YAML and JSON for POCs and MVPs. It's easy and fun. My remarks above relate to production, scale and enterprise needs. The world is a complex place and the variety of approaches available reflects the variety of challenges and business needs. #KnowledgeGraphs #Ontology

AI teams are hitting the same four semantic failure modes that ontologists debated for 25 years — usually without knowing the map already exists.

There is a pattern gaining traction that deserves a direct challenge. The argument goes something like this: model your manufacturing operation using your own naming conventions, store the data in a graph database, and build relationships between the nodes to represent how your assets, processes, an

Over the past few quarters, we’ve seen tremendous growth in enterprise adoption of agentic workflows. Large enterprises are realizing that GraphRAG delivers more accurate results than traditional RAG, multi-hop reasoning is essential for many high-value use cases, and memory and… Read more →

Why the persistent knowledge layer your agents need comes from librarians, ontologists, and database engineers — not from cognitive science or computer science alone

Despite all the use cases and examples of great LLM stuff I come across, I’ve always regressed back to the only two use cases I’m confident for the models:

Answers to questions is not the same as building shared, curated, and auditable knowledge.

"You will be in the top 1% of Enterprise Architects" if you truly understand this one idea: EA is not separate business, application, data, and technology architectures. It is one connected enterprise traceability system. Strategy → Value → Capability → Process → Application → Technology → Transformation Once you see how every layer connects and influences the next, architecture stops being static documentation and becomes a real enterprise decision-making system. Most Enterprise Architecture diagrams explain only a slice of the enterprise. This poster attempts to connect the entire EA in one view: WHY → WHAT → HOW → WITH WHAT → WHEN From: • Stakeholders, drivers, assessments, goals, outcomes, value To: • Capabilities, value streams, resources, courses of action Then: • Business services, processes, actors, objects Into: • Applications, services, components, interfaces Down to: • Technology services, nodes, devices, system software, networks And finally: • Work packages, deliverables, plateaus, and transformation states The most important realization while building this: EA is not multiple disconnected architectures. It is one continuous traceability system: Strategy → Operating Model → Solution Architecture → Technology → Transformation A few deep observations: • Value Streams explain how value progresses. • Capabilities explain what the enterprise must be good at. • Business Processes explain how work executes. • Applications operationalize capabilities. • Technology enables execution. • Implementation & Migration explain how the enterprise changes over time. The real power of EA starts when every layer is traceable to the layer above and below it. This is where architecture stops being documentation and becomes an enterprise decision system. #EnterpriseArchitecture #ArchiMate #BusinessArchitecture #SolutionArchitecture #TOGAF #DigitalTransformation #OperatingModel #CapabilityMapping #ValueStreams #Architecture #EA #BusinessTransformation #ITStrategy | 66 comments on LinkedIn

The Living Graph, Part I: A Genealogy of Intent

Federated Time-Series Queries with maplib's chrontext magic

Learn how graph-on-relational engines work in 2026: PuppyGraph, SQL:2023, ISO GQL, hyperscalers, and Prometheux. A deep dive into the category.

If you have been following the enterprise AI debate, you will have encountered a growing consensus: AI only works when anchored in structured context. Ontology, semantic layers, knowledge graphs — the terminology varies, but the argument is consistent.

I presented my paper “Giving AI a Shared Language: Unified Knowledge Graphs for Multimodal Data” at ICMLT 2026 in Berlin. In previous posts, I explained the method behind it: different data types can be represented in a shared space, and relationships can act as the structure that connects them. I applied this idea to country-level data: country background text from the CIA World Factbook, and time-series indicators from the World Bank. Country borders became the relational backbone. The interesting part was the outliers. Some neighboring countries were economically very different. Some border neighbors had surprisingly different Factbook narratives. Some regions showed uneven neighborhood patterns. And this connects to the bigger question from my last slide: as multimodal AI becomes more powerful, the relationship-aware layer becomes more important — not less. AI can process the pieces. Relationships turn them into connected knowledge. #AI #GraphAI #KnowledgeGraphs #MultimodalAI #RelationshipAwareAI #EnterpriseAI #DataStrategy

Twelve useful questions that all ontologists should ask

AI solutions need to be grounded in an organization’s context. Taxonomies and ontologies can do this heavy-lifting.

When to Use a Labeled Property Graph and When to Use an RDF Ontology-Based Knowledge Graph and when might you need both?

We've talked about Glean's Enterprise Knowledge Graph since 2019. People kept asking to see it, so we built a demo. Ask Glean something like: "show me my manager's projects and the team behind them" and watch it traverse the graph in real time, mapping work and people across your org. Because these entities are pre-indexed in the graph, Glean retrieves the answer instantly. This is what enables multi-hop reasoning and saves the LLM from burning tokens re-reading dozens of documents to piece together the answer on every run. This is just a tiny slice of the data the Enterprise Graph contains. Props to David Huynh for putting this beautiful demo together! | 30 comments on LinkedIn

Less than a month ago I released Ontosphere — a browser-based OWL ontology editor with MCP support, built so that AI agents can build and query knowledge graphs entirely client-side, without a server or a plugin. This post is about what has happened since.