@itsxiaozeng: This prompt hack boosted my AI output quality by ~90%. Most people use lazy role prompts like “Act as a McKinsey consultant” or “Act as a senior software engineer.” The problem? ChatGPT doesn’t know which frameworks, books, methods, or heuristics that expert actually uses—so you get generic, shallow answers (and more hallucinations). Use this 2-step prompt engineering workflow instead: Step 1 — Build the expert persona (resume first). Ask the model to generate a detailed resume for the role: seniority, domain, education, certifications, tool stack, frameworks (MECE, Jobs-To-Be-Done, Porter’s Five Forces), methodologies, typical problems solved, writing style, and decision principles. Step 2 — Activate the persona. Then say: “Act as this person from the resume and [solve X / write Y / design Z]. Show steps, call out assumptions, and state confidence.” Why it works: you’re giving the LLM a concrete mental model, domain language, and constraints—so it stops guessing and starts reasoning like a specialist. This applies to strategy, software architecture, copywriting, finance, marketing, and product—consistently deeper, more structured outputs from any LLM (ChatGPT, Claude, Gemini). Stop vague role prompts. Define the expert first, then deploy it.

Xiao Zeng

Open In TikTok:

Region: US

Wednesday 08 October 2025 00:32:28 GMT