After a dozen-plus questions, the AI stopped and said: let me write up your report.
One line in the report kept me staring at the screen for a long while:
You are not chasing a life you don't have.
The life you want, you are already living it, more or less.
It is rare for a 50-year-old to be pressed into his chair by a single sentence.
This starts with Dan Koe. Early this year he wrote "How to Fix Your Entire Life in One Day", timed with X's million-dollar prize for long-form writers, and it spread widely. I read it years ago, and read it again this year.
But what really hooked me was not that essay — it was another thing of his called Human 3.0, a model of life.
I wrote a prompt based on it and sent it to the AI, asking it to question me. It didn't dump a batch at once; it asked one question at a time. Around the thirteenth, I said that was enough.

01 | It doesn't ask whether you work hard enough — only which subject you are lopsided in
Human 3.0 splits life into four parts: mind (cognition, thinking, emotions), body (sleep, health, energy), spirit (values, relationships, sense of meaning), and career (work, income, freedom).
Why is this split useful? Because we habitually try to solve everything from one quadrant. When work goes badly, the first instinct is to quit; when the heart feels empty, the first instinct is to buy two psychology books.
You cannot fix loneliness by working harder, nor fix being broke with meditation and the law of attraction. First you must know which subject the problem is in.
When my report came out, all four items were at the second layer. No subject was terrible; none was presentable either.
But it did not stop at the platitude "everything is average". It first asked me to recall: your ideal day — did it actually happen at any point in the last 30 days?
I listed six items; five had happened: dinner with family, pushing work forward with colleagues, chatting and watching a movie after a workout, work moving along. The only one that hadn't: a day out with family.
So it said: this is not "something wrong with your life" — it is "inputs and goals not matching". What you truly lack is only two things: contiguous time for experience, and results on the account.
For a 50-year-old, the scariest thing is not a collapsed life. It is being busy every day yet unable to name what you are actually missing.
02 | I have stayed on the second layer for many years
The model also divides each subject into three layers.
The first layer: others tell you how to live — study hard, work hard, marry and buy a home; socialization tasks are handed down one by one. The second layer: you must decide how to live yourself — earn money, build a career, work out. The third layer is not more grinding; it is balance — the freedom, money, experiences and stability you want are all coordinated within one plan.
The second layer has a well-hidden trap: you turn yourself into a project that can never finish being optimized.
All four items on the second layer — that is me.
I get to decide what I do — carrying four lines alone, an autonomy rare among my peers. But the AI pointed out the price: not a single line has a clear numeric standard.
I described my account as "the numbers aren't great, but there's some progress".
It said: this sentence contains two judgments and not a single number.
That is exactly what it looks like. No alarm, no applause. Busy every day, yet nothing can answer "has it actually worked or not".
03 | Getting it and feeling flat doesn't mean you are ungrateful
One more thing — I don't know if you are like this too.
Something you wanted badly before — once you actually get it, it doesn't feel that happy, that exciting.
I used to think this was called ingratitude, and was even embarrassed to admit it. Human 3.0 gives it another name: misalignment — you are not greedy; the old way of living can no longer hold who you have become.
So change is necessary. Not "nice to change things up" — necessary.

04 | The three sharpest cuts in the report
The first: the phone time between 11 and 12.
Six hours of phone a day sounds alarming, but most of it goes to self-media and learning AI — it didn't scold me for that. It picked out only one stretch: after lying down at 11, I sometimes scroll past midnight. It said that is my only obvious leak.
However "justified" the daytime scrolling is, it cannot offset that stretch after lying down.
The second: one action, judged by two sets of standards.
While writing, judge it by the standard of "journaling", written for myself; after publishing, judge it by the standard of "the account", watching the numbers. Then both ends feel wrong.
One action serving two standards serves neither well.
The third cut the deepest. It asked: before you write, is there a specific person in your head?
I said sure, a cohort: 30 to 50 years old, urban, educated.
It said: this definition excludes no one, therefore it does nothing while you write.
Later a specific name came to me: Lao Yang, 50, worried at the time about his son's college and their father-son relationship.
But what was I writing for him? AI, frontier news, mid-life insights.
The answer he wanted and the goods I shipped never matched at all.
The biggest lever it gave me was not working twice as hard, nor learning yet another method —
it was to let one specific person sit across your writing desk, long-term.
The move costs nothing: before writing, put down one line with three blanks — who is this for, what happened to them yesterday, what do they most want to know now.
If the three blanks can't be filled, don't write the piece yet.
Over 30 days you will accumulate a list of people. It said this list says more about whom you write for than the follower count does.
05 | Where AI sits in this tool: not thinking for you, but pressing the questions for you
I am not going to wrap this piece up with three takeaways.
This check-up report could be done with pen and paper; it's just that no one asks in such detail. Here AI is not thinking for you — it does the one thing you cannot do yourself: press on, question by question, neither letting you off nor flattering you.
It doesn't ask "where do you think you can improve"; it asks "where did that hour between 11 and 12 go". Specific to the event, you cannot bluff.
If you want this prompt, leave a comment — just type "want it". I'll reply to each one and send it to you as-is.
Once you have it, give yourself half an hour tonight and send it to the AI. Don't rush to conclusions; the value is entirely in the questioning — many people only discover at round three that it isn't that they don't know what they want; they have simply never been asked seriously.
When you're done, come back and tell me: which question stopped you.
The model comes from Dan Koe's Human 3.0.
AI can take the questions to a place you could never reach on your own.
But the answers only you can give.