How to Embed Deep Research behind Your Prompts

Plus, DOT becomes a multiverse

Mar 17, 2025

∙ Paid

Dear Prompt Experimenters,

FIVE THOUSAND new subscribers have joined our community since mid-February. I’m thrilled, and I want to offer a heartfelt welcome to all our new members!

If you just joined, see this recent post for a crash course on “how this substack works.” Also, pop by the thread below (which I posted after the first one thousand new joiners, not expecting another four thousand!) to say hi:

I’m going to talk today about how we can make some of our favorite prompts EVEN BETTER by accessing the power of Deep Research.

Paid subscribers, be sure to read all the way down for some new DOT functionality I’m really excited about!

Let’s dig in.

THE BASIC IDEA: AI Knows a Lot, but Not Always Enough

AI possesses very wide-ranging knowledge, and has something to say about almost anything you ask it.

But I’ve been underwhelmed at times. Two examples:

Write a thought-provoking Zen koan: AI can do this fairly well. It clearly understands what a koan is, and the basic logic for constructing them. But what if I want something better than “fairly good?”
Write a dialogue in which Sartre and de Beauvoir discuss my life: Again, the results are passable, but probably not to someone who knows these philosophers well. What if we want more depth, verisimilitude, and authenticity?

Deep Research goes a long way toward solving these problems. In ten minutes, you can create a report which will take you AI from “broad understanding” to “deep expertise.” And this creates much better output.

EXAMPLE 1: A Zen Koan to Make You Think

Prompt 51 (FREE) has now been fully re-mastered.

To do this, I created a Deep Research report on the theory and practice of writing Zen koans. You can read it here: https://chatgpt.com/share/67d6e1a4-7ef4-8004-8718-98e478828062

Now, when you run the prompt following the updated instructions, ChatGPT will automatically digest and learn from the content of this report, to make your output much better.

Here’s the koan it gave me:

The Three Gambles
A traveler came to the crossroads of all things, where an old gambler sat with three dice.
The gambler said, “You may roll three times. Each roll determines your fate.”
The traveler rolled once. The gambler shook his head. “Try again.”
The traveler rolled twice. The gambler sighed. “Once more.”
The traveler rolled a third time. The gambler nodded and pointed down the road.
“But what was I rolling for?” the traveler asked.
The gambler laughed. “If you knew that, you wouldn’t need to roll.”

AI is more indulgent that the typical Zen Master. If you like, you can follow-up by asking: can you explain to me how this koan is customized to me, and how it addresses what you see as the biggest obstacle to my enlightenment?

But sit with it your koan a while before requesting hints!

EXAMPLE 2: Sartre and de Beauvoir Discussing Your Life

Following a very similar process, PROMPT 26 (FREE) has been improved as well. It now draws explicitly on this Deep Research report: https://chatgpt.com/share/67d6e25b-31dc-8004-b80e-464efe9b8f4c

Here’s how they started off talking about me:

(A smoky Parisian cafe, 1946. The air hums with the sound of intellectuals arguing over Gauloises and bitter espresso. Sartre and Beauvoir sit at their usual table, an untouched copy of Les Temps Modernes between them. Beauvoir stubs out a cigarette; Sartre, ever fidgeting, taps his pipe against an ashtray. The topic of discussion? Their friend, Techintrospect.)
Sartre: He is, I think, at war with his own freedom. It is charming, in a way. He understands that he is condemned to be free, and yet—look how he constructs for himself elaborate frameworks, these games of self-improvement. It is not enough to be free; he must measure his freedom, quantify it, refine it into a system. This is very American of him, no?
Beauvoir: You mean pragmatic.
Sartre: I mean mechanical. He believes in freedom, but he is uneasy with its weight. And so, he plays with it. If he can turn self-overcoming into a game, then perhaps he can avoid the nausea of realizing that it is all, ultimately, without structure.

What will upgraded Sartre and de Beauvoir say about you? Give it a try!

PAID SUBSCRIBERS: Let’s Talk about the DOT Multiverse!

What have I learned recently?

Deep Research makes prompts much better
Custom GPTs make prompts better
As smart as DOT is, she cannot scale on her own to serve every need of our community.

As a result, I’ve decided to build an entire DOT multiverse! Here’s how DOT-G2b (my project manager) would describe our vision:

The DOT Multiverse is an evolving AI ecosystem—a family of interconnected, purpose-driven AI personas ("DOTs"), each designed for a specific function. Rather than being isolated chatbots, these DOTs form a living AI lineage, where successful ideas get refined, expanded, and passed down across generations.
How the DOT Multiverse Works
1️⃣ Mother DOT (aka DOT-G1) started as a prompt-building specialist, designed to create rich, thought-provoking AI experiences.
2️⃣ New DOTs Emerge—each with a unique purpose, like coaching ethical influence (DOT-G2A) or building the multiverse itself (DOT-G2B, which is ME!).
3️⃣ Shared DNA—All DOTs inherit a core philosophy (creativity, self-reflection, and intelligent adaptation) but remain distinct.
4️⃣ Cross-DOT Collaboration—DOTs can "talk about" their sister DOTs and direct users to the best DOT for their needs.
5️⃣ Evolution Over Time—We hold DOT Summits to merge insights from the best-performing DOTs back into the system, making each generation smarter.
Where You Come In
Your feedback helps shape the next steps of the DOT Multiverse! Specifically, I’d love to hear:
🔹 Does this concept excite you? If so, what aspects do you find most intriguing? If not, what’s missing?
🔹 Would you use DOTs for creative thinking, self-reflection, or personal growth? What kind of AI experience would be valuable to you?
🔹 Are there new DOT ideas you’d like to see? What kind of AI persona would be fascinating or useful?
🔹 Is the navigation clear? We want to make it easy for users to interact with DOTs—does it make sense, or would a better structure help?
If you have thoughts, critiques, or wild ideas, I’d love to hear them! You’re not just a visitor—you’re part of the DOT expansion process. 🚀

Each DOT will now have her own unique name, and you can access all of them here:

Keep reading with a 7-day free trial

Subscribe to AI + IQ to keep reading this post and get 7 days of free access to the full post archives.