Forget prompt engineering: ‘Loop engineering’ is all the rage now AI developers are moving away from direct prompt engineering towards ‘loop engineering,’ where AI agents prompt other agents to complete tasks autonomously. This shift allows users to design recurring systems that guide AI agents, reducing the need for constant manual prompting. While this approach enhances efficiency, particularly in coding, concerns about cost due to increased token usage need to be managed.

• AI creators like Boris Cherny and Peter Steinberger advocate for ‘loop engineering’ over traditional prompt writing.
• Loops are recurring systems that guide AI agents, automating tasks and reducing user intervention.
• An example is instructing an AI to work on a task until completion (/goal) rather than prompting at each step.
• Loop engineering involves components like automations, worktrees, skills, plugins, connectors, and sub-agents.
• A common setup involves one agent writing code and another reviewing it to ensure quality.
• Loops can be applied to various roles beyond coding, such as executive assistants or customer service agents.
• The primary concern with loops is the potential cost increase due to higher token consumption.
• Strategies to manage costs include optimizing API call frequency and using sub-agents only when a second opinion is valuable. Continue reading https://www.businessinsider.com/what-are-loops-ai-engineering-tips-2026-6