A deep dive into AI harnesses, explaining how to build a programmatic environment around an LLM agent to ensure reliability without prompt engineering. The talk demonstrates building a harness for a browser agent to reliably log in and upvote a post on Hacker News using GPT-3.5 Turbo.
Jay Wengrow, author of “A Common-Sense Guide to AI Engineering,” breaks down how AI agents work, describing the 'clever hack' of intercepting LLM output to trigger functions. The discussion covers multi-agent architectures for complex tasks, implementing guardrails with regex and judge LLMs, and a pragmatic take on when to use frameworks versus building from scratch. Wengrow emphasizes understanding fundamentals over specific tools to create robust, production-ready AI applications.
A deep dive into building sophisticated voice agents using OpenAI's Realtime API and Agents SDK. The session covers architectural patterns like chained vs. end-to-end models, the use of multi-agent systems with handoffs for specialized tasks, and best practices for production including debugging with traces, implementing guardrails, and creating robust evaluations.