beyond contect graph : memory for better decisions and learning

book : https://leanpub.com/time-aware-ai-memory article : https://sovereign-agents.leaflet.pub/3mdi3rdyvns2f # Beyond Context Graphs: Why Agent Memory Needs to Be More Than Just a Graph Context graphs are getting a lot of attention for capturing agent decision traces. But here's the reality: a single graph isn't enough for agents to truly learn from experience. **What we actually need:** 🧠 **Episodic Memory with Temporal Causality** Not just "what happened when" but "what caused what, and why." Temporal awareness gives us chronology. Temporal causality gives us understanding. ⚙️ **Cognitive Processing Pipelines** Like human memory, agent experiences need stages: capture → processing → consolidation → reconstruction. Learning happens in the transformation, not just the storage. 🤝 **Promise Graphs, Not Just Action Logs** Mark Burgess had it right with promise theory. Track what agents commit to doing, not just what they did. The gap between promise and action is where the learning