HORMA
The next retrieval gain may come less from searching harder and more from organising memory better.
Signal summary
HORMA is one of the strongest research anchors for HKE's memory direction. It argues for hierarchical memory navigation: agents perform better and use fewer tokens when memory is organised into navigable structures rather than searched as a flat pile of past material.
Why it matters
Many AI systems treat retrieval as a matching problem. HORMA suggests retrieval is also an organisation problem. The system has to decide what structure the past should have, how summaries relate to underlying evidence, and how an agent should descend through that structure for the current task.
LGI reading
This maps directly onto the HKE Foundry idea: a layer that organises knowledge without becoming the canonical record. HKE can use hierarchical organisation to improve navigation while preserving provenance and review gates.
Use this if you are thinking about
memory architecture, source collections, token efficiency, long-horizon agents, temporal errors, and retrieval over living knowledge bases.
Related LGI concept
HKE Foundry
Related signals