The era of long prompts is over? The man behind Claude Code explains how to talk to AI

Boris Cherny, the developer of Claude Code, published a post claiming that the era of long and detailed prompts is over, and that one should talk to AI like a colleague at work. According to him, due to improved models, there is no need to dictate detailed instructions, but rather to convey a goal, level of effort, and method of verification. The post sparked reactions, including criticism of the instruction 'use many tokens'.

Boris Cherny, the man who developed Claude Code, published a post on X this week proposing a simpler approach to working with AI: talk to the model like a colleague at work, instead of formulating long and detailed prompts. According to him, due to the significant improvement in models, there is no longer a need to dictate precise instructions, steps, and limitations. Instead, three main things should be conveyed to the model: what exactly you want it to do, how much effort to invest in the task, and how to verify that the result is correct. The post was written following surprised reactions he received after sharing a prompt he himself used to build an interactive website. The approach contradicts the 'prompt engineering' culture that developed at the beginning of the GenAI era. However, many commenters criticized Cherny for including the instruction 'use many tokens' in his prompt, raising the question of whether the simple approach is also suitable for regular users, for whom a detailed prompt might actually save tokens and unnecessary rounds.