Anthropic's team, the creators of Claude, has published two frameworks to improve communication with artificial intelligences like ChatGPT or Gemini. The first, with five steps, aims to obtain more reliable and task-appropriate responses. It focuses on defining roles, providing context, giving detailed instructions, and including examples—a practical guide for users seeking precision without complications.
Five-step structure for complex tasks 🤖
The proposed technique begins by describing the AI's role and the task in one or two sentences. Then, dynamic content such as texts or links is added for context. The third step is detailed instructions; the fourth, optional examples that clarify the expected format. Finally, critical instructions are repeated in long prompts to avoid deviations. This method is useful for moderately complex tasks, such as document analysis or generating structured reports.
In other words, before we were speaking to it in hieroglyphics 😅
According to Anthropic, until now we asked the AI to juggle data while explaining the task in half a sentence. The novelty is that if you tell Claude it's a finance expert and then paste a 200-page PDF, it responds better than if you just say: make me a summary. Who would have thought, treating the machine like a human with clear instructions works. Good thing someone clarified it before we all ended up ordering pizza from the AI.