Knowledge factory in practice: background agents now 20-50% of all agent runs
hwchase17 · x · 2026-11-22
After LangChain founder Harrison Chase's "knowledge factory" essay, Elias Lumer shared internal agent practices, which Chase amplified.
Overlapping with LangChain's approach:
- Multiple context types — skills, system prompts, user memory, global memory — mostly updated autonomously by background agents
- Background/"dreaming" agents and automations now account for 20-50% of total runs, with intervals from every 1 or 7 days down to as short as 5 minutes
- Data retention shapes just-in-time knowledge pointers: data read from an MCP may only be accessible for 7 days, so it must be saved into user memory/wiki for later
Where they differ: they treat the curator agent and the main agent as the same agent, rather than two separate roles.
More from coding & agent
- Browser dragon-in-your-hand demo built with Three.js, MediaPipe hand tracking and WebRTC — techartist_ · 2026-11-22
- Developer Praises Google Antigravity's Agent Swarm UX — majidmanzarpour · 2026-11-22
- Store Scraper MCP exposes App Store and Google Play data to AI agents — modelcontextprotocol · 2026-11-22
- etapa MCP offers a free AI cycling coach for beginners — modelcontextprotocol · 2026-11-22
- Claude Code plugin turns hooks into persistent memory shared across Codex and Cursor — blaizedsouza · 2026-11-22
- Brandon Galang: As Models Get Capable, UX Scaffolding Melts Away in Favor of Context Flow — brandongalang · 2026-11-22