LangChain CEO Harrison Chase unveils "knowledge factory" architecture for compounding enterprise AI knowledge
hwchase17 · x · 2026-11-22
LangChain founder Harrison Chase shared his new article "How to build a knowledge factory," arguing the missing piece for enterprise AI isn't better models or more context, but better knowledge: systematically collecting and curating the tacit know-how of how a company actually operates.
The accompanying diagram shows a concrete architecture: a single server process (local or on Modal) hosting chat + board, a git repo, a supervisor that starts/restarts agents, and a model gateway tracking costs — with each agent (Claude Code/Codex clones) running in isolated sandboxes. All messages, pushes, and model calls route through the server; agents never talk directly to each other.
Chase calls the pattern a "knowledge factory": brainstorm a mission, then let a swarm of agents autonomously pursue it with monitoring and steering, locally or in the cloud. LangChain's internal GTM and content agents plus LangSmith Engine all use a similar learning loop.
Related event: LangChain Founder: What Enterprise AI Lacks Is a "Knowledge Factory"(17 posts)→
More from coding & agent
- LangChain CEO: enterprises need a 'knowledge factory', not better models — hwchase17 · 2026-11-22
- Harrison Chase: buying Copilot won't transform teams without a knowledge factory — hwchase17 · 2026-11-22
- Aiden Bai doubles down: "** agents!!!" — aidenybai · 2026-11-22
- Cognition CEO Aiden Bai says he never reads the code, just watches agent test videos — aidenybai · 2026-11-22
- Long-running agents fail by re-litigating old decisions, not by crashing — ClickOk5811 · 2026-11-22
- NBER study of 500k+ GitHub devs: coding agents boosted commits ~240% but releases only ~30% — Antique_Atmosphere97 · 2026-11-22