Context Layers, From First Principles

Every vendor in this category now calls some part of its product a "context layer" or an "ontology." The word has stopped telling buyers anything, because each vendor defines it, grades itself against that definition, and declares victory.

So instead of writing a louder version of that claim, we showed our work to someone who checks it for a living. Jessica Talisman is an independent ontology consultant who checks vendor claims about context layers--the glossary, the ontology, the knowledge graph--against the formal definition for each, not just the marketing version. We gave her access to Collate and OpenMetadata product documentation, an open line to CTO Harsha Chintalapani, and agreed to have no editorial control over her conclusions. Her article, "Context Layers From First Principles," runs on her own Substack, Intentional Arrangement. Showing your work only means something if someone else gets to check it.

Why Collate Decided to Show Its Work Instead of Just Claiming to Provide a Context Layer

Anyone can make the claim, "we have a context layer!" based on how they define it. Showing the work means exposing the parts that are hard to fake to someone with deep technical and pragmatic knowledge across all three pieces of it:

1. A glossary that holds a real controlled vocabulary instead of just labels

2. An ontology with typed relationships and logical axioms instead of a hierarchy

3. A knowledge graph that's actually populated and queryable instead of a diagram

My methodology is simple: take the published description of the tool or platform capabilities, compare it against the formal criteria for an ontology, typed classes, typed relationships, axioms that support inference, and see what remains. Most claims reduce to a glossary with a hierarchy at best." — Jessica Talisman, Intentional Arrangement

The ontology test is the sharpest part of her methodology, but it's one piece of a larger question: does the whole context layer, glossary, ontology, and knowledge graph together, hold up against the formal definition of each, not just the version in the product marketing?

What Checking Our Work Means: Typed Relationships and Logical Axioms

Jessica's piece traces the distinction down to Collate's JSON Schema foundation and its W3C standards implementation (RDF, OWL, SKOS, SHACL), and lands on a line of ours she chose to anchor the piece around:

The ontology is the blueprint and the knowledge graph is the building.

A blueprint without a building is just a diagram. Hand someone a building with no blueprint, and all they can do is trust that it won't collapse. Showing the work means handing someone both pieces and letting them check that one actually follows from the other.

What to Go Deeper on Context Layer First Principles?

Jessica is joining Harsha Chintalapani for a live session on October 14 at 8am PT: The Agent-Ready Context Layer.

We'll go past what fits in an article to discuss things like how the ontology and knowledge graph actually work together, what SPARQL and SHACL support looks like in practice, and where the category's definitions still need pressure-testing. We'll follow up after with a short video answering whatever questions we don't get to live.

Read the full piece on Jessica Talisman's Substack.
Register for the October 14 live session.