# How to Build an AI Agent in 30 Minutes Explore the process of developing AI agents with John Berryman. - **Source:** [O'Reilly Radar](https://www.oreilly.com/radar) - **Published:** 2026-07-20 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/41cd02ff-91ef-409e-9530-dbf3f6b212f1 ## What is an AI Agent? John Berryman defines an AI agent as a large language model wrapped in two loops. The outer loop handles message exchange with the user, while the inner loop allows the model to select and run tools, turning it into an agent. ## Natural Language Programming The advantage of programming in natural language is that experts can read the AI's instructions and examine faulty responses. This allows them to make updates directly instead of relaying feedback to engineers. ## AI Agent Development Process John Berryman explains the AI agent development process using the 'outside in' method. He suggests building the traditional software first and then adding the AI component, ensuring the system works without AI. ## Start with a Simple Agent Before writing any AI code, connect the interface to a stub that returns a static response. Then replace the stub with a real but minimal agent. This ensures the model stays on track. ## Agent's Toolkit John Berryman recommends starting the agent with four basic capabilities: read, write, edit, and shell access. This small toolkit allows the agent to extend its own capabilities when necessary. ## Future of AI Agents John Berryman predicts that AI agents will move toward fewer purpose-built applications and more agents that work across tools and interfaces on behalf of individuals, indicating broader use cases for AI. ## Sources covering this story - https://www.oreilly.com/radar/zero-to-agent-in-30-minutes-build-a-workflow-agent-with-john-berryman/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._