The MAD Podcast with Matt Turck Everything Gets Rebuilt: The New AI Agent Stack | Harrison Chase, LangChain
332 snips
Mar 12, 2026 
Harrison Chase, co-founder and CEO of LangChain, a leader in agent tooling and infrastructure. He walks through why the AI stack is being rebuilt: harnesses that manage tools, subagents and files, planning and context compaction, memory types, sandboxes for secure code execution, and observability to run stateful agents reliably. Short, technical and future-focused conversation on the new primitives reshaping autonomous AI.
AI Snips
Chapters
Transcript
Episode notes
Four Types Of Agent Memory And How To Store Them
- Memory splits into short-term (thread), semantic (RAG-like facts), episodic (past conversations), and procedural (instructions/skills).
- In Deep Agents procedural memory is files the agent can update, enabling the agent to 'learn' by editing its own instructions.
Differentiation Comes From Instructions And Tools Not Agent Count
- Whether to build one mega-agent or many subagents depends on use case; the enduring assets are precise instructions, tools, and skills.
- Chase advises enterprises to focus on building instructions and tools because those are portable across architectures.
Invest In Observability Sandboxes And Evals
- Stable, investable infra includes observability, evals, sandboxes, and long-running stateful deployments rather than high-level harness formats.
- LangChain focuses on low-level primitives so products remain useful as higher-level scaffolding evolves.