Writing
Notes on memory for LLM apps.
Design essays from the Memorg team — not announcements, not roadmaps. Just the engineering reasoning behind how the library works.
-
Your chat app's memory is a 50-line bandage on RAG
Most chat-app memory layers are a chat-history table plus a vector store plus a rolling summary. That stack works until it doesn't. Here is why a small dedicated memory layer beats the bandage at the seam where they break.
designrag -
Forgetting as a feature: aging memory in LLM agents
An agent that remembers everything is an agent that remembers nothing useful. Here is why a good memory layer needs to forget on purpose, and how Memorg's importance and recency scoring make forgetting tractable without explicit deletes.
designagents -
Designing memory for agents vs memory for chatbots
Agents and chatbots both need memory, but the shape of what they need is different. Here is how the read patterns, write rates, and scoping requirements diverge — and why Memorg's hierarchy was chosen to serve both without forcing one to wear the other's clothes.
designagents