Harrison Chase writes that LangChain already runs knowledge factories behind its internal GTM and content agents, and that LangSmith Engine, its agent that reviews traces and suggests fixes, maintains one per agent. He says the company is now working with a few customers to bring the approach to their agents. Chase argues the missing piece for company AI is not better models but curated knowledge of how the firm works. He says agents should flag missing or wrong knowledge, and a background curator agent should update it. He adds that no standard has emerged for representing semantic memory, and advises starting with one recurring workflow.

AI summary of the linked source — not editorial content.

Explore the topic

Loading related coverage…

Comments

Loading comments…