本文由 AI 分析生成
建立時間： 2026-03-28 來源： https://x.com/Khazix0918/status/2011640679030800785
Summary
A practitioner shares six prompt “mindset techniques” (心法) for getting better results from AI, emphasizing meta-strategies over rigid templates. The techniques include: letting AI choose its own expert role, using role specificity (Steve Jobs > “10-year PM”), providing structured context before asking, iterative refinement, and other conversational tactics learned from two to three years of daily AI use.
作者分享六個 Prompt 使用心法，強調思維技巧而非固定模板。技巧包括：讓 AI 自行選擇最適合的專家角色、角色具體化（賈伯斯 > 「10 年 PM」）、先提供結構化上下文再提問、迭代精煉等，來自兩三年日常使用 AI 的實踐經驗。
Key Points
- Let AI choose its own expert role: specify the domain and problem type, then ask AI to select the best expert persona before answering
- Specific historical/real people as roles outperform generic titles (“Steve Jobs” > “10-year product manager”)
- Express context and background clearly before asking — half the battle is accurate problem description
- Templates are less important than the underlying skill of clear communication and problem articulation
- When unsure who the ideal expert is, let AI ask 2 clarifying questions before choosing a role
Insights
The technique of letting the AI choose its own role is a form of meta-prompting that offloads the “persona engineering” problem to the model itself. Specific real people work better than generic roles because they carry rich, trained associations — the model has a detailed model of how Steve Jobs thought, vs. a vague prototype of “product manager.” The repeated emphasis on “expressing the problem clearly” underscores that prompt engineering is primarily communication skills, not magic incantations.
Connections
Raw Excerpt
你能把问题表达清楚、把上下文和背景信息表达清楚，你的Prompt就成功了一大半。