I'm working on a new retrieval system. Not RAG

It uses TCF ( Temporal Cognitive Fields) to create CFGS ( Cognitive Field Geometry Shapes) for persistent, stateful recall.
The strongest property of this architecture is continuity. Because retrieval is stateful and tied to episodes, themes, persistent facts, and live field geometry, the system is better positioned than standard RAG to resume ongoing work, preserve relationship context, and adapt response planning to current internal state.
The main trade-off is complexity. A system that retrieves across multiple memory forms and can reactivate them into a live field is more powerful, but it also needs stronger discipline around speaker attribution, telemetry isolation, and continuity hygiene. The architecture gains expressive control at the cost of a more demanding runtime and more subtle failure modes.
Not sure if I'm releasing the source code yet, but I'm honestly thinking I should. Project is here.: https://AuraCoreCF.github.io