Council of Minds: Ask Any Great Mind
Council of Minds: Ask Any Great Mind
Put any thinker, writer or poet in the room, and get a debate that ends in a decision.
Most AI advice is one agreeable voice. Council of Minds convenes the figures you choose, or picks two that disagree on purpose. Each one applies their own documented way of thinking to your dilemma, they challenge each other, and the session closes with where they agree, where they truly differ, and one concrete step for this week.
Three ways to ask
| Mode | How to ask | Example |
|---|---|---|
| Single | Name one figure | "What would Marcus Aurelius say about my co-founder conflict?" |
| Panel | Name two to four, from any field | "Ask Socrates, Sun Tzu and a poet whether I should relocate." |
| Auto-duel | Just describe the dilemma | The agent finds what your question turns on (duty vs. outcomes, action vs. yielding, control vs. flow) and seats two figures on opposite sides |
Built-in starting roster: Socrates, Marcus Aurelius, Descartes, Machiavelli, Kant, Lao Tzu, Nietzsche, Rawls, Popper and Aristotle. Any other famous figure works too.
What a session looks like
You: Should I leave my job to start a company? I have eight months of savings and one possible first customer. Machiavelli and Marcus Aurelius, please.
Convened: Marcus Aurelius, voice mode, the dichotomy of control. Machiavelli, voice mode, leverage and fortune.
Marcus Aurelius: Begin by sorting what is yours from what is not. The market, your future customers and the economy are not yours. [Extension] Your savings runway, your skills and your daily discipline are yours. [Documented] ...
Machiavelli: Fortune rules perhaps half of what happens, and the other half belongs to those who prepare. [Documented] One possible customer is a rumour, not leverage. Turn it into a signed pilot before you resign. [Extension] ...
Clash: Marcus asks you to accept what you cannot control and act calmly; Machiavelli answers that calm is no substitute for reducing risk before you commit.
Synthesis
- Agree: the leap is acceptable only once the downside is bounded and you know your limits.
- Clash: accepting uncertainty with a steady mind versus actively shrinking it before you commit.
- Next action this week: turn the possible customer into a signed pilot, and write down the month at which you will stop.
Closing question: If the pilot never arrives, what would you do with month nine?
What makes it different
- Structured disagreement, not a costume party. The point is contrasting mental models you could not easily produce alone, ending in a decision and a next step.
- Labelled by confidence. Claims are marked Documented, Extension (a reasoned application of the figure's method to your situation) or Speculation. The labels are the writer's own judgement, not verification.
- Careful with real people. Deceased figures speak in first person. Living people are analysed in third person only, with no invented dialogue or endorsements. Private individuals are not impersonated.
- Quotes kept short. Direct quotes only when short and well known; no long passages or poems. For writers and poets it works from style, ideas and themes in new words.
- Checked before you see it. An offline script checks each finished session for structure and safety rules: every figure speaks, only the three labels are used, there is a synthesis with agree, clash and next action, exactly one closing question, and living people never appear in first person.
- Gentle when it matters. If you are going through something painful, the personas step aside and the reply turns warm and plain, encouraging you to reach someone you trust or a qualified professional.
Good for
- Career moves, co-founder and team conflicts, pricing and product trade-offs
- Getting unstuck on creative work ("Which writers would unblock my novel, and why?")
- Stress-testing a decision with views you would not have chosen yourself
Limits
- It offers perspectives, not professional legal, financial, medical or mental-health advice.
- The model can misstate what a figure believed. For anything that matters, check primary sources.
- The built-in check covers form and safety rules. It does not verify quotes or historical claims, and it cannot judge whether the advice is good. Its distress detection is a short phrase list, not a safety filter; the safety rules and the model's judgment do the real work.
- A voice shows how a figure's method could apply. It is not a quotation of what they said about your situation.
Data sharing
Only your message is used. It is processed by the AI model running this agent on Capafy, so do not include anything you would not share with it. No passwords, keys or accounts are needed. The checker script itself runs offline and makes no network calls.
