Agent Factory Recap: Agent Harnesses, Shifting Left, and Autonomous Coding
Google, Friday, September 25th, 2026
Google's Agent Factory podcast recap explains what an 'agent harness' is and how grounding models in context enables greater agent autonomy.
This recap of a Google Cloud 'Agent Factory' episode features engineer Ryan Lopopolo, who coined the term 'agent harness,' discussing what it takes to build effective autonomous coding agents.
It defines an AI agent as an LLM plus a harness, the surrounding tooling that supplies live context, such as querying weather data before answering whether to bring a raincoat.
Lopopolo discusses shifting interventions earlier in a workflow and treating team collaboration processes as a way to unlock higher levels of agent autonomy. The post links to timestamped segments of the full conversation.