LangChain's 'knowledge factory': the curator agent is what enterprises are missing
hwchase17 · x · 2026-11-22
LangChain founder Harrison Chase published 'How to build a knowledge factory', arguing the missing piece for enterprise AI isn't better models or just more context — it's better knowledge of how the company actually works.
- The key concept is a curator agent: knowledge shouldn't be maintained manually (too much work) or left to the consuming agents (they flag issues well but fixing requires deep understanding; self-updating leads to unruly, contradictory knowledge).
- Instead, a background curator agent should run on a cadence, with access to raw data sources (Notion, Slack), agent traces (to codify usage patterns into skills), and the issues agents have flagged.
- nichochar calls these space-attached agents 'gardeners', vs person-attached 'butlers' (OpenClaw, Grok Bot, Muse). Chase notes LangChain's internal GTM/content agents and LangSmith Engine all share similar learning loops.
More from coding & agent
- Charting the Singularity: a Claude-powered world model run on 60K+ prompts — MauiHawk · 2026-11-22
- LangChain's Harrison Chase: Companies Need a 'Knowledge Factory,' Not Better Models — hwchase17 · 2026-11-22
- Beyond AI Coding: Developers Should Move to the Five-Layer Agentic Engineering Stack — Pavan_Belagatti · 2026-11-22
- PSAISuite Pipes Native PowerShell Cmdlets Into Multi-Model Agent Harness — dfinke · 2026-11-22
- Testing shows renaming a Git checkout doesn't change MemoraX's memory scope ID — Elegant-Chance-6420 · 2026-11-22
- Lean 4 formalization of Masullo's Riemann proof hits 51%: 1213 of 2385 theorems proved — bremen79 · 2026-11-22