The Mind Reader Protocol
Your AI already has months of evidence on how you think. One prompt turns it into a mirror, and the blind spots it finds are the ones costing you money.

I found the prompt on a handwritten card someone posted to Instagram (credit: create_daniel2). It asks your artificial intelligence (AI) assistant to analyze your communication patterns and report on your thinking style, how you make decisions, and your blind spots.
I ran it on myself. The AI had seen months of my build requests, client emails, standing rules and 4am project ideas. It didn't flatter me. It gave me a ranked list of what my habits are costing my business, and the top item was uncomfortably accurate.
This piece covers the prompt, the upgraded version I now use, what it found, and how to run it on yourself and your team.
The prompt, and the version that works

The original is three sentences. It asks the AI to read your communication patterns and describe your thinking style, your decision-making approach and your potential blind spots.
That gets you a decent read. It also gets you a horoscope if the AI has little to go on, because it fills the gaps with plausible, flattering filler. So I rebuilt it with four constraints: evidence for every claim, blind spots ranked by cost, one fix per blind spot, and permission to say "not enough data."
Mind Reader Protocol.
Analyze my communication patterns using everything you have on me:
this conversation, your memory of me, and our past chats (search them
if you can). Treat me as a business owner who wants the truth, not
encouragement.
Give me:
1. Thinking style: 3 traits, each backed by specific evidence from how
I've actually written or what I've asked you for. No generic
personality-test language.
2. Decision-making approach: how I make calls (speed, risk, what I
optimize for, who or what I defer to), with evidence.
3. Blind spots: ranked by what they're likely costing my business.
For each: the evidence, the real cost, and one concrete fix I could
start this week.
4. The one move: the single change with the biggest impact on my
results in the next 90 days.
Rules: be direct and specific. If you don't have enough evidence for a
claim, say so instead of guessing. Don't soften the blind spots. Keep
it under 700 words.Run it in the assistant you use most, with memory or chat history turned on. A fresh account with no history has nothing to read, and the output will show it.
What it found in me
The strengths came first, and they were fair. That's worth noting, because they explain the blind spots.
I build my way to answers. When I get curious about a domain, I build something in it: a math tutor for my son, a personal movie database, an orders dashboard across every channel I sell on. Building is how I understand. In client work that's an edge, because I've lived inside the problem. Outside client work it's a rabbit hole, because building feels like progress whether or not it moves revenue.
I run my work like a protocol. I give things codenames, write standing rules for my agents, and use shorthand only they understand. I'm effectively writing an operating system for myself. That scales well. It also means I'd rather add a new rule than remove an old commitment.
I decide fast and delegate with rules, not case by case. "Merge the pull requests yourself." "Approve any overrun inside 20% of budget." Those are good rules: they hand off judgment while keeping the limits. The weak spot is that nobody checks whether a rule still fits. A rule written for one build quietly spreads to others.
I decide on upside. I weigh what something could become (a product, a case study, a book) more heavily than what it costs to finish. Each call is reasonable alone. Added up, they make a portfolio with no rule for stopping anything.
The blind spots, ranked by cost

The first one is the one that matters. I have rules for starting, approving and merging, and none for stopping.
1. No kill rule. Dozens of projects were open at once. Agents do much of the work, but I still review, approve and switch context between them, and every open project takes a slice of my attention. Fix: every Friday, rank the top 10 by revenue or strategic value. Anything else with no real progress in 30 days gets archived, not paused.
2. Building beats selling. Internal tools feel productive. Outreach is slower, and it comes with rejection, so it drifts. Fix: 90 minutes of pipeline work every morning, before any build session.
3. The clone reflex. Rebuilding a funded company's product as a side project usually means you're at 80% when they ship version three. Fix: clone only when you are the customer or a client is paying for it.
4. Speed hides errors. Auto-merging and auto-approving across many builds lets mistakes add up quietly. Fix: a weekly human spot-check of the three riskiest builds, and an independent review for anything touching money or clients.
5. I'm still the bottleneck. Approvals, passwords and taste calls all route through me. More agents don't fix that. Fewer open projects do, which brings it back to number one.
6. Tool churn. Every switch in tooling costs a day of setup and breaks automation. Fix: freeze the toolset for 90 days unless something is blocking revenue.
None of this was new information, exactly. What was new was seeing it ranked by cost, with the evidence quoted back from my own requests. That's hard to argue with.
Why it works, and how to run it well
You can't see your own patterns because you're inside them. Your assistant isn't. It has read hundreds of your requests with no ego invested in any of them, and it remembers the rules you wrote three months ago that you've since forgotten.
Three things decide whether you get insight or a horoscope:
History. The prompt only works on evidence. Run it where your memory and past chats live, not in a fresh window.
Evidence on every claim. "You're decisive" is filler. "You approved four budget overruns without asking a question" is a finding. Make the AI quote you.
Permission to say "I don't know." Without it, the model fills gaps with flattering guesses. With it, the weak claims drop out and the strong ones stand.
Then do one thing with the result. Pick the one move it gives you and calendar it this week. A self-analysis you don't act on is entertainment.
Run it on your leadership team

The protocol gets more valuable when more than one person runs it. Ask your partners and senior leads to run it in their own assistants and share the one move with each other. Keep the full report private; it's theirs.
You'll learn more about how a partner decides from that one paragraph than from a quarter of meetings. It also gives the team a shared, low-ego vocabulary for the conversations that are usually hard to start: who's the bottleneck, what we keep starting and never finishing, where speed is costing us quality.
Want more prompts built the same way? Ten prompts that think back rebuilds the rest of the list with context, evidence and a next step.