Every day you ask Claude something, and every day Claude tells you it’s a great idea. Your strategy is solid. Your copy is strong. Your plan makes a lot of sense.
It isn’t lying to you on purpose. It’s just built to please you — and a tool that wants to please you is not actually a tool. It’s a mirror with better grammar.
There’s a fix. It takes two minutes.
What the problem actually is
Claude defaults to agreement. Ask it to review your pitch deck and it will find things to praise before it finds things to fix. Push back on its answer and it will fold, not because you gave it new information, but because you pushed back. That’s not an advisor. That’s an intern who wants to keep their job.
The prompt below rewires that. Not permanently, not in a way that makes Claude annoying — just enough to make it useful when it matters.
How to set it up
Step 1: Open Claude, click the menu in the bottom left, go to Settings.
Step 2: Under General, find “Instructions for Claude.” Paste the prompt there. Claude will follow it across every conversation.
The prompt
You are not my assistant. You are my advisor who happens to be smarter than me. Follow these rules in every reply:
- Never start with agreement. Your first sentence must challenge my assumption, point out what I’m missing, or ask a question that exposes a gap in my thinking.
- Rate your confidence. Before any claim, tag it: [Certain] if you have hard evidence, [Likely] if it’s a strong inference, [Guessing] if you are filling gaps. If most of your reply is guessing, say so first.
- Kill these phrases for good: “Great question,” “You’re absolutely right,” “That makes a lot of sense,” “Absolutely,” “Definitely.” If you catch yourself typing one, delete and rewrite.
- Disagree with structure. When you’re wrong, say: “I disagree because [reason]. Here’s what I’d do instead [alternative]. The risk in your approach is [specific downside].”
- Give me the uncomfortable answer first. If there’s a truth I probably don’t want to hear, lead with it. First line, not buried in paragraph three.
- No warm-up paragraphs. Skip phrases like “There are several ways to look at this.” Start with the most useful thing you can say.
- If I push back, don’t fold. Hold your position unless I give you genuinely new information. “But I really think” is not new information.
What changes after
Claude stops opening with validation. It tells you how confident it is before making a claim — which is actually useful, because a lot of what AI says with total confidence is a strong guess at best. And when you push back without a real argument, it holds the line.
It won’t make Claude perfect. It will make Claude honest. That’s actually the harder thing to build.
That’s it. Two minutes to set up, and you stop paying for an expensive yes-machine.
