up until yesterday we relied on our own reason stack, now we have the new OSS model which I am layering in as another model the ai can pull if it wants another thought outside of its own. I think once we get closer to a beta after things are polished on the blackwell we will probably start to demo off some of that. As to the memory store and all its fun stuff with vectors and the likes I would love to show, but we have our own concepts. V9 right now uses a 4096 dim model just for one aspect of its memory. V8-V8.1 is a much less dense model but its not trying to be like V9. its more a product agent system but with the learning. V9 is where we are playing in the fun spaces of research where we get things like this which we spend hours, days , weeks tracing the why it feels that way.
@dmitryrichard I have better traceability than just logs 
Neo4j (native graph)
This will just give you small insight into the power of the systems I am playing with vs what a lot of people are building and this will also give insights on how we designed some of the memory systems