The New AI Trend: Loop Engineering Is Taking Over from Prompt Engineering

Prompt engineering may be giving way to a more advanced concept—"loop engineering". Boris Cherny, founder of Claude Code, recently told CNBC bluntly: "I no longer write prompts. Now an agent prompts Claude." He emphasized that the loop mechanism is what makes this shift possible, and predicts that such technology will become one of the accomplishments he's most proud of in a decade.

This trend is not an isolated case. Peter Steinberger, an OpenAI engineer and creator of the viral project OpenClaw, publicly reminded users on X: "This is your monthly reminder—you should no longer manually write prompts for coding agents." Claire Vo, founder of ChatPRD and podcast host, summed it up: "It's really just reminding people that you don't need to type prompts with your own fingers—the agent can do it for you."

From "writing prompts" to "designing loops," AI interaction is shifting from one-shot instructions to automated iterative closed-loop systems. Loop engineering enables agents to self-drive and repeatedly call on models, greatly reducing human intervention. As this concept spreads, developers are expected to focus more on building efficient workflows rather than fine-tuning prompt text, thereby unlocking AI's autonomous capabilities.