The visual breakdown of RAG vs Agentic RAG vs Memory is incredibly clear. I particularly appreciate the distinction between 'read-only via tool calls' and 'read-write via tool calls', that shift from static retrieval to active memory management is the real unlock for personalized agents. One challenge I've seen with long-term memory implementation is the 'memory pollution' problem you hinted at. Just as we need garbage collection in programming, agents need semantic garbage collection to prune outdated context, otherwise retreival precision degrades over time. It's interesting to see databases like Agentic Postgres tackling the forking branching aspect of this state management.
Hi Avi..I hope you are well. I have e been reading a few of your posts and they are very intriguing. I just want to know exactly where, I know some RAG flows you ha e mentioned it being done on Figma, do you make your diagrams/Rag flows etc..is there a way we could discuss or you could help? Would be very interested in knowing more
This is great, incredibly clear designed post!
The visual breakdown of RAG vs Agentic RAG vs Memory is incredibly clear. I particularly appreciate the distinction between 'read-only via tool calls' and 'read-write via tool calls', that shift from static retrieval to active memory management is the real unlock for personalized agents. One challenge I've seen with long-term memory implementation is the 'memory pollution' problem you hinted at. Just as we need garbage collection in programming, agents need semantic garbage collection to prune outdated context, otherwise retreival precision degrades over time. It's interesting to see databases like Agentic Postgres tackling the forking branching aspect of this state management.
Hi Avi..I hope you are well. I have e been reading a few of your posts and they are very intriguing. I just want to know exactly where, I know some RAG flows you ha e mentioned it being done on Figma, do you make your diagrams/Rag flows etc..is there a way we could discuss or you could help? Would be very interested in knowing more