Let Your AI Agent Do ITS OWN Harness Engineering (Here's How)
Harness engineering is how we make AI agents better over time: improving the rules, skills and memory around the model every time the agent gets something wrong. Most of us still do that by hand. In this video I show you how to let your agent do its own harness engineering, learning from real conversations and proposing changes to its own harness, while you stay in the driver's seat and approve everything it learns. It's the same playbook that makes coding agents reliable, applied to any agent you run in production, with LangChain's Managed Deep Agents and LangSmith Engine. ~~~~~~~~~~~~~~~~~~~~~~~~~~ - Try Managed Deep Agents with $50 in LangSmith credit (use code ColeM): https://www.langchain.com/referral/cole-medin/?utm_medium=creator&utm_source=youtube&utm_campaign=q3-2026_cole-medin-dedicated-1_aw Thanks to LangChain for working with me on this video! ~~~~~~~~~~~~~~~~~~~~~~~~~~ - Join Dynamous for the new Agentic Coding Course 2.0 - the core of the course is now COMPLETE: https://dynamous.ai - Managed Deep Agents docs: https://docs.langchain.com/langsmith/python/managed-deep-agents-overview - LangSmith Engine: https://docs.langchain.com/langsmith/engine ~~~~~~~~~~~~~~~~~~~~~~~~~~ 0:00 Harness Engineering Beyond Coding Agents 0:55 Production Agents Are Built Like Coding Agents 1:26 The Harness, Mapped to Production Agents 2:01 Two Loops for Improving Your Agent 2:39 Managed Deep Agents 3:54 Building the Agent 5:33 Loop 1: Memory from Escalations 7:20 Where the Lessons Live 7:48 Live Demo: Teaching It a Lesson 10:23 Same Problem, Different Member 10:59 Private Memory per User 11:34 Loop 2: LangSmith Engine 13:27 Staying in Control 14:16 Versions, Rollbacks, and Slack 14:47 The Results 15:18 Final Thoughts ~~~~~~~~~~~~~~~~~~~~~~~~~~ Join me as I push the limits of what is possible with AI. I'll be uploading videos weekly - at least every Wednesday at 7:00 PM CDT!
