Stanford CS329Z Lecture 7: How AI Agent Memory Works — MemGPT & Memory Types
zainhas · x · 2026-11-16
Stanford's CS329Z "Engineering AI Agents" course released Lecture 7 by Michael Ryan on how AI agent memory works, covering MemGPT and memory types. The recap table distills four patterns: learning procedures from experience (reusable code/workflows; reuse saves effort but goes stale), tiered memory hierarchy (compact active memory + external detail; requires reliable management), reflective memory consolidation (lessons and higher-level inferences; abstraction aids reuse but risks unsupported conclusions), and memory maintenance (revised facts, indexes, representations; better organization costs compute and can lose info). Video is on Stanford Online.
More from coding & agent
- Track cost per finished task, not token price: the agent metric that matters — arpit_bhayani · 2026-11-16
- Qwen4 Max lands in OpenCode Go, Alibaba Qwen team confirms — Alibaba_Qwen · 2026-11-16
- Procedural generation via Claude-written scripts: copyright-free photobash assets — Promptmethus · 2026-11-16
- Agentic harnesses hit a plateau: "the ceiling on intelligence is the model" — alexisgallagher · 2026-11-16
- QuixiVoice: a Rust voice stack for any agent, from VAD to TTS in one pipeline — QuixiAI · 2026-11-16
- Music video made with Claude Code driving Blender: physics-solver dolphin animation — maximstark_ · 2026-11-16