# Stop Prompting Claude Code, Start Engineering Loops: Master Agentic Automation > Source: > Published: 2026-07-15 15:01:04+00:00 Member-only story # Stop Prompting Claude Code, Start Engineering Loops: Master Agentic Automation ## A prompt is a request. A loop is a system In early June 2026, a tweet from Peter Steinberger, the developer behind the OpenClaw framework, hit five million views in under a day. The gist of it: developers should stop prompting coding agents one message at a time and start designing loops that prompt the agents instead. Around the same time, Boris Cherny, who leads Claude Code at Anthropic, said something similar in his own words: he doesn’t prompt Claude anymore, he has loops running that do it for him. His job, he put it, is to “write loops.” That’s a strange thing to hear from the person who built the tool. But it captures something real that happened to Claude Code over the last several months: it quietly shipped a full set of native looping primitives, and most people using it every day have never touched any of them. They’re still typing a prompt, waiting, reading the response, and typing the next one, one turn at a time, like it’s 2024. ## A prompt is a request. A loop is a system. The moment you stop asking Claude Code to do one thing and start defining the condition under which it should stop doing many things, you’ve crossed from operating the tool to engineering with it.