What is: Agentic Looping?

Stop writing single prompts and start designing autonomous agentic loops. In this video, Quinton Wall breaks down the core architecture behind AI agents like Claude Code, inspired by creator Boris Cherny’s philosophy: "My job is to write loops." Learn how the ReAct (Reason + Act) loop works, why most teams fail when deploying agentic workflows, and the 3 non-negotiable code-level guardrails you need to prevent runaway token spend and infinite loops. Key Takeaways: • Why prompt engineering is evolving into loop architecture • The ReAct framework: Reason, Act, Observe, and Repeat • The 3 core components of an agentic loop in code • Why stop conditions must live in code; not prompts • 3 essential guardrails: Iteration caps, progress detectors, and token budget ceilings Learn More: • Explore Postman AI Tools: https://postman.com 0:00 - Boris Cherny: "My Job is to Write Loops" 0:15 - What is an Agentic Loop? 0:32 - The ReAct Framework (Reason, Act, Observe, Repeat) 0:43 - 3 Core Components of an Agent in Code 0:56 - Why AI Loops Fail (The Runaway Budget Problem) 1:12 - Prompts are Suggestions, Code is Law 1:18 - 3 Essential Guardrails for Agentic Loops 1:32 - Shifting from Prompter to Systems Architect