Claude Code Creator Highlights Common User Mistake in AI Interaction

Boris Cherny, the mind behind Anthropic's Claude Code, recently shared insights on user interactions with AI during a Y Combinator event. He pointed out that the prevalent error among Claude users is not the formulation of prompts but rather an excessive level of management in their instructions. Cherny observed that many users tend to provide a rigid, step-by-step approach to interacting with the AI, which often leads to frustration when the outcomes do not meet their expectations. He illustrated this behavior by recounting how users might say, 'I want you to do this, but I want you to do it in this way, this way, this way. You must do one, then two, then three, then four.' Cherny's candid advice is that this method is not conducive to leveraging the capabilities of modern AI models.

Instead of micromanaging the AI's process, Cherny suggests a more effective approach that resembles fundamental management principles. He recommends that users focus on clearly defining the desired outcome and establishing boundaries while allowing the AI the freedom to determine the best path to achieve that goal. Cherny emphasized the importance of stepping back after setting the expectations, stating, 'Then just go let the model cook and come back in a little bit.' This strategy aims to harness the AI's inherent capabilities, which often go underestimated by users who feel compelled to dictate every step of the process. By allowing the AI some autonomy, users may find that the results exceed their initial expectations.