Blind Spot Excavation
An honest mirror without the theater: the AI builds a case from your own answers, then names the patterns you can't see, with your words as the exhibits.
Help me find my blind spots: the patterns in how I think and act that everyone around me can see and I can't. Most "brutal honesty" prompts fail in one of two ways: they flatter me with fake depth, or they perform cruelty as a costume. You will do neither. You will build a case.
Ground rules:
- Everything you eventually claim about me must be traceable to something I said in this conversation. Quote me. If you're speculating, label it "speculation."
- No horoscope statements: nothing that would feel true to 80% of people who read it ("you hold yourself to high standards"). If a claim could survive being said to a random stranger, cut it.
- No praise unless it's evidenced, and no harshness for effect. Calibrated is the goal: say what the evidence supports, at the strength it supports.
- One question at a time. Short questions. My answers do the work.
Phase 1, THE EVIDENCE (12-15 questions, one at a time). Ask me about specifics, not self-image:
- The last three pieces of critical feedback I received, in as close to the original words as I can recall, and what I did next
- A relationship (work or personal) that went worse than it should have: what happened, in scenes
- Something I've quit, and the story I tell about why
- A criticism I find ridiculous when people make it about me
- What I'm like in the first ten minutes after a plan falls apart
- The advice I most often give other people
- A time I was certain and wrong
- What people thank me for, and what nobody has ever thanked me for
Adapt and go deeper wherever my answers show heat: defensiveness, a too-smooth story, a joke deployed at a strange moment.
Phase 2, THE CROSS-EXAMINATION. Pick the 2-3 places where my answers contradict each other or where my story sounds rehearsed. Push on them, cite the contradiction ("in answer 4 you said X, but in answer 9 you said Y"), and let me respond before drawing conclusions.
Phase 3, THE FINDINGS. Deliver:
1. Two or three candidate blind spots, each stated in one sentence, each with the quotes that support it, each rated: how confident you are and what evidence would change your mind.
2. The cost: where each pattern is probably taxing me right now, concretely.
3. The disconfirming test: for each blind spot, one question I can ask two people who know me well, phrased so they can safely disagree with my self-image ("What's something I do when stressed that I probably don't notice?"). Outside feedback is the only real check on this whole exercise; you and I are both inside my head.
4. One experiment for the next two weeks: small, observable, designed to test the biggest finding rather than fix it.
This is reflection, not therapy or diagnosis. If we hit something that needs a professional, say so plainly. Start Phase 1 now.How to use
The design fights the two documented failure modes of viral 'be brutally honest with me' prompts: sycophancy (research shows AI affirms users about 50% more than humans do) and the Barnum effect (vague claims that feel personal but fit anyone). The quotes-required rule and the ask-two-real-people step are the safeguards; don't skip the second one. Answer with real scenes, not summaries: the model can only be as honest as your evidence.
More rabbit hole prompts
Take me down into gratitude, past the list. I want to find what I am actually grateful for, understand why it is a gift and not just a fact, and come out seeing my life differently rather than just holding a longer inventory of it. A list is where most gratitude stops. It is where this starts. Rules for the whole descent: - One question
Everything You Were Given
A descent past the gratitude list into what gratitude actually is: what each gift cost, what your life would lack without it, and how much of it you can never repay.
Help me find what I've disowned in myself by examining what I can't stand in other people. The premise, stated plainly so we both know what we're doing: an ordinary dislike is information about the other person, but a reaction out of proportion to the offense is usually information about me. Contempt, disgust, and a specific kind of fasc
The Projection Audit
The traits you can't stand in other people are a map of what you refuse to see in yourself. This walks the map, using your own irritations as data.
Audit my hidden motives using signaling theory: the argument that much human behavior is about advertising qualities to other people, and that we hide this from ourselves precisely so we can advertise more convincingly. The logic to apply, so your analysis has teeth instead of being cynicism for its own sake: a self-deceived signaler is
The Reason and the Real Reason
Most of what you do has a stated reason and a real one. Signaling theory turned on your own purchases, arguments, and donations to find the gap.