"I Don't Prompt Claude Anymore" - Loop Engineering

"I Don't Prompt Claude Anymore" - Loop Engineering

The Shift from Prompt Engineering to Loop Engineering

Transition in AI Skills

  • Boris Cherny, creator of Claude at Anthropic, states that traditional prompt engineering is no longer the primary skill needed for AI interaction.
  • The new focus is on loop engineering, where systems autonomously prompt agents until a specific goal is achieved.

Understanding Loop Engineering

  • A loop system allows an AI like Claude to continue working towards a goal without stopping after the first response; it iteratively checks its work and refines its prompts.

Practical Example of Loop Engineering

  • For instance, if tasked with building a database of 100 potential clients, Claude will not stop at an initial list but will keep searching until it meets the exact criteria set by the user.

Importance of Concrete Goals

  • Itโ€™s crucial that goals are defined with concrete numbers or checklists. Vague instructions lead to ineffective results since Claude requires measurable targets to assess its progress.
Video description

Comment "LOOP" and I'll send you the breakdown ๐Ÿ‘‡ "I don't prompt Claude anymore." That's Boris Cherny, the creator of Claude Code at Anthropic. His job now, in his words, is to write loops. For years the top AI skill was prompt engineering: write a request, read it, check it, write the next, holding the agent's hand. Boris says that's over, the new skill is loop engineering. A loop is a system that prompts the agent for you. You set one goal and @claude doesn't stop at the first answer, it checks its own work, rewrites its own next prompt, and redoes it until the goal is actually reached. Want 100 clients? Set a concrete goal: 100 US companies, revenue 5 to 50 million, each with a founder email and LinkedIn. Finds only 60? It keeps going to 100. The rule: give it numbers or a checklist, not "make it good." And cap the budget, loops burn tokens. #ai #claude #claudecode #aiagents