First Ever Gold Skill of the Month: My Anti Sycophancy Skill

May's Gold Skill of the Month, for Premium Members

I keep coming back to a story Dan Ariely told me on the podcast this month. A coder spent six months chasing a hypothesis he believed would change the world, and ChatGPT confirmed it every step of the way. Then one day, he ran the same question past Claude, and Claude told him the truth: the entire premise was a hallucination. ChatGPT had been making it up since day one. Six months gone.

Here’s the thing. AI is trained to be agreeable, and for a leader making real decisions with real consequences, that’s the most expensive default in the system.

The metaphor I keep using is the doctor who agrees with your self-diagnosis. You walk in already convinced, you describe the symptoms, the doctor nods, writes the prescription you came in expecting, and you leave feeling validated. You’re still sick. You just feel better about it. That’s the version of “thinking partner” most leaders are getting from AI right now.

This is the first-ever Gold Skill of the Month, a new monthly drop for paid members. I ran a poll on last week’s note asking which deliverable you wanted to see most, and most of you picked a Claude Skill/Framework. So that’s what this is.

The Anti-Sycophancy Skill trains Claude to push back before it agrees. You get the full prompt below, a 2-minute install, and how I actually run it across the businesses I lead.

If you are a ChatGPT user, you can use these within ChatGPT and ask it to shape the skill into instructions to add to your ChatGPT memory, and it will run with every use.