A feature article discusses why writing prompts for AI can consume significant time and effort, sometimes producing prompts longer than the answers themselves, while focused teams still devote effort to crafting the right words to set a large language model (LLM) in motion, with some teams employing dedicated prompt engineers sometimes more schooled in creative writing than computer science. Noting that creating good prompts remains an evolving art, the article presents a list of 20 different prompt styles and structures to use alone or combined for a hybrid strategy. Among them, instruction-based prompting offers a concise but detailed step-by-step description of what the LLM should do, with length, tone, and structure spelled out for consistent results, though instructions can be longer than the answer. Few-shot or example-based prompting gives the LLM a few examples, useful for formats like JSON or CSV. Zero-shot prompting suits cases where examples aren’t necessary or feasible, such as translating a paragraph into French or Russian. Q&A format treats the LLM as a sage, using short questions followed by direct answers to triangulate on an answer, working when the LLM is a tool for finding knowledge.