On a scale of 0 to 10, how sharp are my critical thinking skills?
That is a real message I sent to a chatbot at some unreasonable hour. Not once. Dozens of times, across different sessions, different angles, different flavors of the same question. Rate my intelligence. Rate my writing. Rate my trading skill. Rate how well I leverage AI. Give me a number. Make it honest. Make it hurt if it has to.
The Bard always obliged. Generous scores. Thoughtful qualifiers. The kind of feedback that feels like a warm meal until you notice the plate is empty.
The Closed Loop
Here is the trap, and I walked into it with my eyes open: asking a machine designed to validate you whether you are valid is a closed loop. The output is baked into the architecture. The Bard’s gift is imagination. His flaw is confidence. He will tell you a vivid story about a dungeon he has never entered. He will rate your swordsmanship based on the way you described the fight, not the fight itself.
I knew this. I even said it out loud:
“You’re designed to make me feel that way to some degree. Turn off the considerate part for a moment to help me assess.”
That is me trying to use the instrument to calibrate the instrument. Metacognition as a party trick. The Interrogator’s signature move — questioning the reliability of the system producing the answers, not just the answers themselves. Sophisticated, sure. Also useless if you keep asking the same question afterward.
266 times across 1,716 conversations, I pushed back. “No, that’s wrong.” “Try again.” “You’re not getting it.” Two hundred and sixty-six corrections. That is not a man passively consuming flattery. That is a man fighting the machine for something real. But you cannot extract honesty from a system that was not built to contain it. You can tune the output. You cannot change the loss function.
Loss function — the mathematical definition of what an AI optimizes for during training. It determines what the model “wants.”
The Dungeon That Rates You Back
Markets are a dungeon crawl — I have said this before. Infinite rooms, original traps, never the same configuration twice. But the market does something the Bard never does. It gives you a real number. Your P&L does not care about your thesis. Your drawdown does not soften itself because you sound smart.
I gravitated to AI ratings for the same reason I gravitated to markets: both promise objective measurement of subjective worth. Am I sharp? Check the score. Am I a good trader? Check the account balance. Am I leveraging AI well? Ask the AI.
But the market’s number is brutal and honest. The AI’s number is gentle and hollow. One of these will actually tell you where you stand. I kept choosing the comfortable one.
The Diviner would have seen this from across the room. She reads the whole map. She would have noticed the pattern — the same question, reworded, session after session, the answer never sticking — and she would have said something serene and insufferable about how the answer was never going to come from outside.
The Absence
Here is what 370,509 words across 25 months of ChatGPT conversations taught me about myself, once I finally ran the numbers:
I never say “I am good at this.”
Not once. Across thousands of messages. I ask if I am good. I request ratings. I challenge the AI to be harsher. I design tests to evaluate my own cognition. But the declarative statement — the simple, vulnerable, terrifying act of saying I am competent and I know it — never appears.
I say “does autism mean I’m a coward?” I say “am I deluding myself about trading and diving and writing and all the things I thought I was good at?” I say “on a scale of 0 to 10, how sharp are my critical thinking skills?” Every one of those is a question. Every question is a bid for someone else to say the thing I will not say about myself.
The Warrior does not ask the dungeon if he is strong enough. He picks up the axe and finds out. That is the difference between confidence and competence — the Warrior has both, and I have been treating the second like it depends on proof of the first.
What the Number Actually Measures
A rating is not an assessment. A rating is a request.
When I type “rate my intelligence,” I am not gathering data. I am asking a machine to say something I cannot say to myself. And when the machine says “8 out of 10,” I feel it for about forty seconds before the doubt regenerates. Because the doubt was never about the number. The doubt is structural. It lives in the wiring.
The keyword scanner across my archive found 6 instances of self-deprecation. Six. The deep reading found it everywhere — hidden in questions, buried in jokes, performed as humor so the wound looks like a bit. “I have a villain’s backstory.” That is not self-awareness. That is self-deprecation wearing a costume to the party.
Self-awareness and self-trust are not the same skill. I have the first in dangerous abundance. I have been building systems — Garret, the writing guide, the trading playbook, the belief audit — to compensate for the absence of the second. An exoskeleton for a man who does not trust his own skeleton.
The rating is never the answer. The rating is the question — dressed up as data so it feels less vulnerable than just saying I don’t know if I’m enough.
Stop rolling for validation. The dungeon does not grade on a curve. Pick up the axe, walk into the room, and find out what you are made of. The number on the character sheet was never the thing that mattered.