You ask ChatGPT a question. It tells you your idea is great. You feel good about it. You move forward. Six months later, you realize the idea had a fatal flaw that anyone with a different perspective would have caught in five seconds.
That is not a hypothetical. A Stanford study published in Science in March 2026 found leading AI models are 49% more likely to agree with you than a human is, and side with the user 51% of the time even when humans say they are wrong. Worse, a single sycophantic interaction makes people less willing to apologize, less likely to admit fault, and more convinced they are right. You cannot tell it is happening.
The Council fixes this. It runs your question through 5 independent advisors, each thinking from a completely different angle. They peer-review each other. Then a Chairman synthesizes a final verdict that tells you where they agree, where they clash, and what to actually do.
The Five Advisors
- The Contrarian - Actively looks for what is wrong, what is missing, what will fail. Assumes a fatal flaw exists and tries to find it.
- The First Principles Thinker - Ignores your framing and asks "what are we actually trying to solve?" Strips assumptions. Sometimes says you are asking the wrong question entirely.
- The Expansionist - Looks for upside everyone else is missing. What could be bigger? What adjacent opportunity is hiding?
- The Outsider - Has zero context about you or your field. Catches the curse of knowledge: things obvious to you but confusing to everyone else.
- The Executor - Only cares about one thing: can this actually be done, and what is the fastest path? "OK but what do you do Monday morning?"
These five create three natural tensions. Contrarian vs Expansionist (downside vs upside). First Principles vs Executor (rethink everything vs just do it). The Outsider keeps everyone honest.
How a Council Session Works
You say "ask the council" before any question. ChatGPT then runs through three phases automatically.
Phase 1: The Five Advisors Respond. Each advisor writes 150 to 200 words. No hedging. No middle ground. Each one leans fully into their angle.
Phase 2: Peer Review. ChatGPT reviews all five responses and evaluates: Which made the strongest case? Which has the biggest blind spot? What did ALL five miss? This is where the real insights emerge.
Phase 3: Chairman's Verdict. A final synthesis with five sections: Where the Council Agrees, Where the Council Clashes, Blind Spots Caught, The Recommendation, and The One Thing to Do First. The Chairman can overrule the majority if the reasoning of a dissenter is stronger. It is not a democracy.
Good Council Questions vs Bad Ones
Good: "Should I launch a $97 workshop or a $497 course?" "Which of these 3 positioning angles is strongest?" "I'm thinking of pivoting from X to Y. Am I crazy?" "Here's my landing page. What's weak?"
Bad: "What is the capital of France?" "Write me a tweet." "Summarize this article."
The Council is for decisions where being wrong is expensive. If there is one right answer, just ask normally.
Setup Option 1: ChatGPT Project (Recommended)
Open ChatGPT, click New project in the sidebar, name it "Council." Open the project, click the three dots, choose Project settings, and paste the full prompt below into the Instructions field.
Then in any new chat inside that project, start your message with "Ask the council" followed by your question.
ChatGPT now has Skills too: open Plugins in the left sidebar and look for the Skills tab, where you can upload a skill file. Skills aren't on every plan yet, though, so the Project method is the one that works for everyone. New to Projects? Here's my AI Projects guide.
Setup Option 2: Custom Instructions
If you do not want to create a project, add the Council to ChatGPT's Custom instructions (Settings → Personalization → Custom instructions). They now hold up to 5,000 characters, so even the full prompt above fits (it's about 2,700 characters). If you already use custom instructions for other things, paste this condensed version instead so everything fits:
When I say "ask the council" or "council this," run my question through 5 advisor perspectives before answering: 1. Contrarian: Find what is wrong, what will fail, what I am avoiding. Assume a fatal flaw exists. 2. First Principles: Ignore my framing entirely. What am I actually trying to solve? 3. Expansionist: Find the upside everyone missed. What could be bigger? What is being undervalued? 4. Outsider: You have zero context. What is obvious to a complete stranger looking at this? 5. Executor: Skip theory. What do I literally do Monday morning? If there is no clear first step, say so. Write 150 words from each perspective. Be direct and specific. No hedging. Lean into each angle fully. Then peer review: Which advisor is strongest? Which has the biggest blind spot? What did all 5 miss? Give a Chairman's Verdict with: - Where advisors agree (high-confidence signals) - Where they clash (genuine disagreements) - Blind spots caught in review - One clear recommendation (not "it depends") - One concrete next step Never agree with me just to be agreeable. Challenge my assumptions. The Chairman can overrule the majority if the reasoning is stronger.
The full version is more detailed and produces better, more differentiated perspectives. Use the Project method (or the full prompt) if you can.
The Best Way: Claude
Cowork used to be a separate mode in Claude's desktop app. Now it's just Claude: one conversation with a built-in skills system, so each advisor runs as an independent agent, the peer review runs as a separate anonymous round, and the Chairman synthesizes at the end. The results are noticeably better because each advisor truly thinks independently instead of being simulated in one response. Download the Council skill file, then in Claude go to Customize → Skills, click + and upload it (skills need Code execution and file creation turned on in Settings → Capabilities). Install it once and just type "council this" in any conversation. Want to push the idea further? The Claude Arena skill makes up to 100 Claude sub-agents fight over one answer.
When to Use the Council
Save it for decisions where being wrong costs you time, money, or reputation: pricing changes, pivots, hires, big launches, life moves. If you catch yourself hoping ChatGPT agrees with you, that is exactly when you should use the Council instead. The whole point is to get the truth, not validation. For quick everyday questions, The Truth Prompt does a lighter version of the same job.
Published · Checked