AI Trends 2025: AI Agents and Multi-Agent Systems with Victor Dibia - #718

This TWIML AI podcast episode interviews Victor Dibia, a Principal Research Software Engineer at Microsoft Research, about AI agent innovations in 2024. The conversation covers Dibia's background, his work on AutoGen (a multi-agent framework), and key trends in AI agents, including increased enterprise adoption (often as thin wrappers around existing APIs), the rise of agent-native foundation models with built-in reasoning capabilities, and the development of interface agents that interact directly with user interfaces like web browsers. Dibia discusses the challenges and opportunities of building complex, autonomous agent systems, highlighting the importance of frameworks like AutoGen for managing complexity and providing tools for building production-ready applications. A key takeaway is the need for specialized models and improved evaluation methods to address the limitations of current agent technology.