The Anthropic leader who built Claude Code says he ditched prompting — now he just writes loops.
The lead developer of Anthropic's Claude Code, a tool for AI-assisted coding, has publicly stated that he no longer uses traditional prompt engineering. Instead, he now writes loops—a technique he calls 'loop engineering.' This approach involves using iterative code structures to guide the AI model's output, rather than crafting detailed natural language prompts. The developer's comments, reported by The New Stack, suggest a fundamental shift in how developers interact with AI coding assistants. By writing loops, the developer aims to achieve more reliable and controlled results from the AI, moving away from the unpredictability of one-shot prompting. The article does not provide specific examples or code snippets of the loop engineering technique, nor does it offer benchmark comparisons or performance metrics. However, the developer's position as the lead of Claude Code gives weight to the claim that this method represents a new paradigm in AI-assisted development. The source text does not include any details about the effectiveness of loop engineering compared to traditional prompting, nor does it mention any official Anthropic endorsement or documentation of the technique.
Loop engineering could change how developers interact with AI coding assistants, moving from prompt crafting to iterative code structures.