Isn't lugging around all that memory "baggage" going to become cumbersome to the models we use? The more memory you bring along, the larger the footprint of what has to be fed into the context window.
Granted, in my mind, this basically just looks like RAGing in memory from model to model, and I may be looking at this over-simplistically. Is there a technique you have in mind that helps streamline the extra context needed?
You’re right dumping all memory into the context window doesn’t scale. But with CORE, we don’t do that.
We use a reified knowledge graph for memory, where:
Each fact is a first-class node (with timestamp, source, certainty, etc.)
- Nodes are typed (Person, Tool, Issue, etc.) and richly linked
- Activity (e.g. a Slack message) is decomposed and connected to relevant context
This structure allows precise subgraph retrieval based on semantic, temporal, or relational filters—so only what’s relevant is pulled into the context window.
It’s not just RAG over documents. It’s graph traversal over structured memory. The model doesn’t carry memory—it queries what it needs.
So yes, the memory problem is real—but reified graphs actually make it tractable.
Isn't lugging around all that memory "baggage" going to become cumbersome to the models we use? The more memory you bring along, the larger the footprint of what has to be fed into the context window.
Granted, in my mind, this basically just looks like RAGing in memory from model to model, and I may be looking at this over-simplistically. Is there a technique you have in mind that helps streamline the extra context needed?
You’re right dumping all memory into the context window doesn’t scale. But with CORE, we don’t do that.
We use a reified knowledge graph for memory, where: Each fact is a first-class node (with timestamp, source, certainty, etc.) - Nodes are typed (Person, Tool, Issue, etc.) and richly linked - Activity (e.g. a Slack message) is decomposed and connected to relevant context
This structure allows precise subgraph retrieval based on semantic, temporal, or relational filters—so only what’s relevant is pulled into the context window. It’s not just RAG over documents. It’s graph traversal over structured memory. The model doesn’t carry memory—it queries what it needs.
So yes, the memory problem is real—but reified graphs actually make it tractable.