The Prompt That Makes Chatgpt Go Cold

Task: [specific task]. Role: [expert persona]. Context: [background]. Constraints: [limits]. Format: [output format]. Quality bar: [standard]. Validate before responding: does your plan meet all requirements?

Run a thought experiment on [topic]. State your assumptions clearly. Walk through the logic step by step. Identify where intuition breaks down. What counterintuitive conclusion emerges? Be rigorous.

How would 3 different experts disagree about [topic]? Expert 1: [type], Expert 2: [type], Expert 3: [type]. Each gives their strongest argument. Then: where does the truth likely lie?

What's the uncomfortable truth about [topic] that most people avoid? State it directly. Then explain: why people avoid it, what the evidence says, and what to do about it. Don't sugarcoat.

Everyone thinks [common belief] about [topic]. Make the strongest case that the opposite is true. Use: evidence, logic, historical precedent, and first principles. Don't be contrarian for the sake of it — be genuinely persuasive.

I was told [advice I received]. Give me a second opinion: Is this good advice? For whom specifically? What context makes it right or wrong? What would you actually do in my situation?

Apply 3 different mental models to [problem]: 1) First principles, 2) Inversion, 3) Opportunity cost. For each: what it reveals, what it misses, and what the combined insight is.

Make 5 specific, falsifiable predictions about [topic/trend] for [timeframe]. For each: probability, key indicator to watch, what would change your mind. Track record: how often are you right?

Problem: [symptom]. Ask 'why' 5 times, drilling to root cause. For each 'why': state the cause, provide evidence. At the root: what's the real fix? Not the band-aid — the cure.

Audit this content for quality: [paste]. Rate 1-10 on: accuracy, originality, depth, clarity, actionability. For each: specific examples of what's strong, what's weak, and exactly how to improve.

What can [field A: e.g., biology] teach [field B: e.g., business]? Identify: 3 successful cross-domain transfers that already exist, 3 that haven't been tried yet, and why the untried ones might work.

Here's my thinking on [topic]: [explain]. What hidden assumption am I making? What if that assumption is wrong? What changes if I drop it? Identify the load-bearing assumption.

Take my plan: [describe]. Now design the exact opposite plan. What would that look like? Is there any scenario where the opposite is actually better? What did I learn?

Make me a 5-minute expert on [topic]. What are the: 3 essential concepts, 2 key debates, 1 thing everyone gets wrong, and 3 questions to ask that signal I know what I'm talking about?

What can't you do well regarding [topic]? Be honest about: your knowledge cutoff, areas of uncertainty, likely biases in your training data, and where a human expert would dramatically outperform you.

I'm deciding between [options]. Help me write a decision journal entry: what I'm deciding, key factors, what I expect to happen, my confidence level, and when I'll revisit. Protect me from hindsight bias.

My plan/code/argument: [describe]. Find the edge cases: extreme scenarios, unlikely combinations, boundary conditions, and failure modes. For each: what happens, how likely it is, and how to handle it.

I received this feedback: [paste]. Help me: separate signal from noise, identify patterns across feedback sources, determine what to act on vs. dismiss, and create specific improvement actions.

Here are my tasks/goals: [list]. Using the Eisenhower matrix (urgent/important), rank everything. For top 3: why they matter most. For bottom 3: why they can wait or be eliminated.

Map my knowledge on [topic]: what I know confidently, what I think I know but might be wrong about, what I don't know I don't know, and the most important thing to learn next.