1/ Since @JayaGup10 and I wrote about context graphs, the response has been huge...but I've also noticed a few misconceptions worth addressing.

The tldr: context graphs aren't a graph database or structured memory. They require a fundamentally different approach to schema and representation.
This matters because I'm seeing teams reach for familiar tools (Neo4j, vector stores, knowledge graphs) and wonder why their agents aren't getting smarter. The primitives are wrong.

A few things I want to clarify: 2/ "Ontology" is an overloaded term. We need to be more precise.

There are prescribed ontologies (rule engines, workflows, governance layers). Palantir built a $50B company on this: a defined layer mapping enterprise data to objects and relationships. You define the schema. You enforce it. It works when you know the structure upfront.

The next $50B company will be built on learned ontologies. Structure that emerges from how work actually happens, not how you designed it to happen. This is important because there’s so much implicit knowledge in decision making that we don’t even realize in the moment, and agents are meant to replicate our judgement!