Context Engineering Our Way to Long-Horizon Agents: LangChain’s Harrison Chase

The discussion centers on the evolution and future of long-horizon agents, particularly in coding and AI Site Reliability Engineering (AISRE). It highlights the shift from simple prompts to complex harnesses that facilitate context engineering through techniques like compaction and sub-agents. A key point is the increasing importance of traces in agent development, serving as a source of truth alongside code and enabling online testing and team collaboration. The conversation explores the role of human judgment in evaluating agent performance, including the use of LLMs as judges calibrated against human preferences. The potential for agents to improve themselves by analyzing traces and modifying their code is also examined, alongside UI considerations like asynchronous management and synchronous communication modes.