How to write AI agent loops, explained

Lenny's Newsletter just published a walkthrough of how to actually build AI agent loops — the kind where a model doesn't just spit out one answer, but keeps thinking, checking its work, and trying again until it's done.

The piece lays out the mechanics: prompt the agent, let it reason, review its output, loop back if something's off. It's the difference between asking a friend for a recommendation and having them research, cross-check, and come back with a real answer.

What's interesting is that these loops are becoming the standard way to get AI to do actual work — not just generate text, but handle decisions, corrections, and multi-step tasks. That shift is why so many teams are moving past simple chatbots toward agents that can run through a process on their own.

Why this matters for us: if you're using AI to get things done, you're probably already running these loops whether you know it or not.