AI’s Next Big Shift: Moving From Prompt Engineering to Loop Engineering

In the world of artificial intelligence, prompt engineering was long considered the essential skill for productive AI interaction. But today, loop engineering is quickly becoming the new favorite among top AI practitioners. According to Business Insider, Boris Cherny — the creator of Claude Code — has openly stated that he no longer writes prompts. Instead, he lets AI agents drive their own work through self‑running loops. In a CNBC interview, Cherny explained that one of his agents sends instructions to Claude on its own, while he focuses purely on building the loop structure. He even believes that loops and their related features will be what he’s most proud of a decade from now.

And Cherny is not alone. OpenAI engineer Peter Steinberger, creator of the viral project OpenClaw, posted a monthly reminder on X telling users they should stop writing prompts for coding agents. Claire Vo, founder of ChatPRD, summed it up by saying the core idea of loop engineering is: You don’t have to type prompts with your own fingers to get an agent working for you. This philosophy is reshaping how AI applications are developed — shifting from manually designing one prompt at a time to building autonomous loop systems that iterate on their own.

Loop engineering marks a leap from “instruction‑based” to “agent‑based” AI interaction. Developers are no longer acting as typists for AI; instead, they become architects of loop systems — defining task goals, feedback mechanisms, and boundary conditions so that AI agents can make decisions, execute, and improve autonomously within the loop. For teams chasing efficiency and scalability, this could mean a massive leap in productivity, though it also raises the bar for system robustness and explainability.