http://www.heyuan110.com/tags/openhands/Recent content in OpenHands on Bruce on AI EngineeringHugoenTue, 08 Sep 2026 10:00:00 +0800http://www.heyuan110.com/posts/ai/2026-09-08-cmu-11-768-ai-agents-course/Tue, 08 Sep 2026 10:00:00 +0800http://www.heyuan110.com/posts/ai/2026-09-08-cmu-11-768-ai-agents-course/<p><img src="http://www.heyuan110.com/posts/ai/2026-09-08-cmu-11-768-ai-agents-course/cover.webp" alt="CMU 11-768 AI Agents Fall 2026 syllabus breakdown: build a harness, evaluate, train with RL" loading="lazy" decoding="async" fetchpriority="low" width="1200" height="630" /> </p> <p>Ten minutes into the second half of lecture 1, Graham Neubig asks the room a question I&rsquo;ve spent most of this year writing about. There are two ways to make an agent better, he says: train the LLM, or engineer the harness around it. Show of hands, which one matters more? A few hands go up for training. A lot go up for harness. Daniel Fried raises his hand twice and gets told instructors don&rsquo;t get to vote.</p>