Meet "loop engineering": The next evolution in AI coding isn't a better prompt, it's a system that prompts itself

Forget typing prompts. A growing number of AI developers are moving beyond one-at-a-time instructions and building what they call "loops" — automated workflows that let AI agents keep working toward a goal without constant human input.

OpenAI engineer Peter Steinberger put it plainly: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents."

Boris Cherny, creator of Claude Code, has already made the shift. He told CNBC, "I don't write the prompt anymore. Claude writes the prompt, and now I'm talking to that new Claude that is kind of coordinating."

The approach borrows from traditional computing — think scheduled tasks and parallel processing — but applies them to autonomous AI systems. Often, one agent writes code while a separate agent reviews it. As Google Cloud's Addy Osmani noted, "The model that wrote the code is way too nice grading its own homework."

The trade-off? More agents mean more compute costs. But for many developers, the efficiency gains are worth it.