LangChain’s Harrison Chase says agents need trace-driven learning and swappable models
hwchase17 · x · 2026-07-21
LangChain’s Harrison Chase responds to Jeff Huber’s “12-factor companies” essay with two core ideas for agent systems: - **Agents should learn from usage.** He argues that traces are the foundation: you need to observe what happened, collect feedback, and update the system over time. - **Own context, rent intelligence.** Companies should retain their tacit and institutional knowledge while keeping models swappable. He notes the phrasing can be confusing, and points to Satya Nadella’s framing: there is **general intelligence** in the models you rent, but each company also creates its **own intelligence** through usage. Overall, he connects the essay to how LangChain thinks about building compound learning systems for agents.
More from coding & agent
- Matt Pocock says every new codebase turns legacy within days — mattpocockuk · 2026-07-21
- Ramp says its model router cuts LLM costs by 30% by picking the cheapest good model — rohanpaul_ai · 2026-07-21
- Meta and Unity link AI workflows to Quest development across setup, input and validation — Vjeux · 2026-07-21
- AI Engineer World’s Fair spotlights Kids Day with 87 children learning to code — steveonjava · 2026-07-21
- Looking Glass adds persistent coding sessions that can schedule their own next turns — teleport66 · 2026-07-21
- Kimi K3 looks stronger and about 5× cheaper on a frontend dashboard task — OwariDa · 2026-07-21