Ray Dalio built an AI clone of himself — it still cannot write your principles

1. The clone answers in his voice. That is the problem, not the feature

On 21 October 2025, Ray Dalio posted the beta of Digital Ray, an AI twin trained on four decades of his writing, principles and decision rules. He put numbers on it himself: about 80% as good as talking to him on markets and economics, around 95% on life and work principles. There is a waitlist.

Say you get in. What do you actually ask it?

Most people would ask it to decide something for them. That is the wrong use — and not because the AI is weak. It is because Dalio has spent fifty years arguing against exactly that. In Principles: Life and Work, the bet is that a principle is a decision rule you extracted from your own failures, and that borrowed principles go quiet at the moment you need them. He published Principles: Your Guided Journal in November 2022, subtitle Create Your Own Principles to Get the Work and Life You Want — roughly three years before he shipped the clone. Notice which one he thought did the real work.

Use Digital Ray to hear a very good imitation of a very good investor. Fine. But your principles have to come out of your mouth, and an AI turns out to be unusually good at pulling them out.

2. Get the source straight: five sections, not five slogans

Part II of Principles: Life and Work — "Life Principles" — is five sections. Every viral version of this topic is a compression of those five, and compression is where the blade gets lost. So start from the original wording.

Section 4 is the one everybody flattens. "Understand yourself and others" reads like a communication problem you can talk your way through. The book says the wiring differs at the level of how a person perceives reality. You don't coach that out of someone. You route around it.

The five-step process inside section 2, in Dalio's own order: have clear goals; identify and don't tolerate the problems standing between you and them; diagnose those problems to their root causes; design a plan that gets around them; push the design through to results. Most of us are excellent at step one and quietly forge step three.

3. Step-by-step: make the AI interview you (prompts you can copy)

Skip the obvious move. Ask for "10 life principles" and you get ten sentences nobody could disagree with — which is the same as none. The interview is where it is won.

1. Hand it raw material before it's allowed to ask anything

An AI asked to interview a stranger interviews a stranger. Give it the mess first: your PrinciplesYou report (free, and it spits out a profile you can paste), or failing that, your last five real decisions written out by hand — what you chose, what you gave up, what you got wrong, how long you took.

Here is my raw material. Read it and use it only to make your questions sharper.
Do not quote it back at me. Do not write any principles yet.

---
[PrinciplesYou profile, or: the last five decisions I made.
For each: what I chose, what I gave up, what I got wrong, how long I took to decide.]
---
Reply with one line only: the single most interesting contradiction you found.

2. Force it to ask one question at a time

This is the whole trick. Models want to close. They summarise early, they flatter, they hand you a tidy document by turn four. You have to forbid it in writing.

You are a strategist who is good at spotting what someone is avoiding.
Interview me about my life principles, using Ray Dalio's five-part structure:
Reality, Goals, Openness, Human Nature, Decision Quality.

Rules:
- Ask ONE question at a time. Never two.
- No summaries. No praise. Never write "that's a great insight".
- If my answer would apply to any person alive, say so in one line and ask the
  same question again, aimed at a specific event in my life.
- You may push back. If I contradict something I said earlier, quote both and ask
  which one is true.
- Do not produce any output until, for each of the five parts, you have at least
  one answer containing a real event, a real number, or a real person's name.
- Do not close before round 12.

Start with part 1. Ask your first question now.

3. Make it go after the answer you dodged

The best question in the whole session is the one you answered badly. Ask for it by name.

Before you write anything else: name the one question I answered worst, and say why.
Then ask it again, harder. If I dodge a second time, write the answer you believe
is true and ask me to confirm or deny it in one word.

4. Turn the answers into principles that can fail

A principle that costs nothing is a preference. A principle nobody could disprove is a bumper sticker. Both fail the test below.

Convert the interview into principles. For each one:
- One sentence, under 25 words, no semicolons.
- Name the cost. If it costs me nothing, delete it.
- Make it falsifiable: state what would prove it wrong in the next 12 months.
- Tag it: Reality / Goals / Openness / Human Nature / Decision Quality.
- Delete any principle that would still be true for a stranger with my job title.
Show me the ones you deleted and why.

5. Compress it to one page and cut the universal ones

Output a one-pager in plain text:
- 10 core principles, numbered, one line each, with the cost in parentheses.
- 5 life lessons I learned the hard way. Each must contain a specific event.
  If a lesson could be printed on a motivational poster, rewrite it.
- 3 decisions I face in the next 90 days, and which principle fires on each.
No heading longer than four words. No bullet poetry. No encouragement.

6. Wire the page into the next real decision

A one-pager nobody consults is a journal entry. Give it a job this week.

Here is the decision in front of me: [one paragraph, with real numbers and dates].
Run my one-pager against it. For each principle that applies, say what it tells me
to do AND what it tells me not to do. Then name the two of my principles that
conflict here — there are always two — and say which one wins and why.

4. What this doesn't replace

The journal is worth naming properly: Dalio built it around four things — identifying your strengths, weaknesses, motivations and preferences; running the five-step process; being assertive and open-minded at the same time; and squeezing the learning out of mistakes. Those four are a better checklist than most AI-generated summaries you'll see this year.

5. Traps, and who this is not for

And the trap I hit myself: the model starts agreeing with you around turn eight. It has learned your voice by then and it wants to be agreeable. When the questions get soft, say so and ask it to attack your last three answers.

6. FAQ

Is Digital Ray free

It launched as a waitlist beta; I don't know what it will cost at general release, and I'd rather say that than guess.

Do I need to read the book first

No. The five-section structure is enough to run the interview. Reading it afterwards is more useful, because you'll recognise which of your principles are really his.

Which model should I use

Whichever one you'll actually sit with for an hour. Long back-and-forth is where they differ more than on any benchmark — some models wrap up and hand you a summary at round six no matter what you wrote in the rules.

What if I hate what it writes

That's usually the useful part. The principles you resist are the ones you don't currently live by — which is exactly why they're worth writing down.

How long does this take

An hour for the interview, twenty minutes for the one-pager. The reruns are fast: once the page exists, step 6 takes two minutes before any real decision.

Will this make me more decisive

It makes you more consistent. Decisive is a separate problem, and I don't think a prompt fixes it.

7. Do this today

Take the PrinciplesYou assessment, paste the profile into prompt 1, and run the interview tonight. Don't aim for ten good principles — aim for three you'd actually obey on a bad Tuesday.

Conclusion: the AI's job here isn't to know you. It's to keep asking until you stop performing.