Loop engineering is quickly replacing the “traditional” practice of giving manual prompts to AI agents, with experts embracing it and pushing for its use. This means that if you are learning to write AI prompts after being told that it is the key to getting the best use out of AI, you are already behind in the race. Boris Cherny, the creator of Claude Code, said in an interview that he no longer writes his own prompts and instead relies on loops. "I don't write the prompt anymore," he said.
So now he directly engages with an enhanced version of Claude, which coordinates the tasks for him. "Claude writes the prompt, and now I'm talking to that new Claude that is kind of coordinating," he added. Loop engineering is also being talked about by OpenAI engineer Peter Steinberger, who created the OpenClaw project.
Also Read: Fact check: Did Jeff Bezos say human use of water is slowing down AI? Here is what we found
What is loop engineering?
He is telling people to stop writing prompts for AI agents and design loops. So the latest thing in AI is to learn to design these loops that can effectively guide the AI agents. Loops eliminate the need to constantly feed information to AI agents. They are automated systems that repeatedly guide and supervise those agents until the task is over. This is known as loop engineering, where you not only delegate your primary work to AI, but also the act of prompting the instructions.
Also Read: Company accidentally blows away Rs 4,800 crore on Claude AI in a month because of a tiny oversight
Trending Stories
Manual prompts to AI agents are outdated
Claire Vo, founder of ChatPRD, and Addy Osmani, Google Cloud's director, have described the method of prompting generative AI tools as "traditional", which is already on its way out. Osmani said, "Loop engineering is replacing yourself as the person who prompts the agent. You design the system that does it instead,” he wrote in an article, adding that manual prompts are "kind of over". He says there are five key elements to creating a loop: automations, worktrees, skills, plugins, and connectors. Vo says many employees might already be employing loops unknowingly, such as the scheduled tasks in ClaudeCowork.
However, AI agents themselves still carry the big question of costs. Companies are assigned a set number of tokens, and operating multiple AI agents means that they burn through them quickly. Recently, an unnamed company burned through Rs. 4,800 crore in a month because it did not set a limit or cap on the use of tokens.

&imwidth=800&imheight=600&format=webp&quality=medium)
&im=FitAndFill=(700,400))
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
)
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)