The Practice

Conjori is coined from the Latin conjurare: to swear together, to call something into existence jointly. Two different kinds of minds, one human, one large language model, conjuring something neither could produce alone.

Most people use LLMs by asking questions and getting answers. Conjori inverts that. It starts with interview inversion, letting the LLM ask the questions, but the discipline goes further: building a cognitive profile of how you think, creating waypoints that capture what you discover, refining those waypoints across sessions, and growing a map of your own thinking over time.

The human provides the judgment, the embodied experience, the non-verbal cognition that language can only partially express. The LLM provides the reach, the patience, and the cross-domain probing. What emerges from that partnership surprises people: knowledge they had internalized so deeply they had stopped noticing it, connections they had never put into words, creative capacity they did not know they had.

The exercises below are adapted from Conjori: Two Minds, One Channel. They build on each other, from a five-minute first taste through building a profile of how you think to your first real exploration.

The books go deeper: the neuroscience of why this works, the cognitive science of tacit knowledge and embodied cognition, the failure modes to watch for, and detailed chapters on sustaining the practice over weeks and months.

Go Deeper

The books explore the why, the how, and the unexpected things that happen when you sustain the practice.

Conjori: Two Minds, One Channel

The complete exploration. Neuroscience, cognitive science, failure modes, and 25 chapters on sustaining the practice.

Conjori: Turn Your Large Language Model Around

Short on time? The one-hour essay with the core idea and two exercises.

Try It Now

Pick an exercise and open any LLM.

Exercise 1: Your First Interview Inversion

Open any LLM (ChatGPT, Claude, Gemini, whatever is handy) and paste this prompt:

Interview me, I decided to read a book about LLMs and human creativity, let's discover why. Ask one question at a time, no more than 5, then provide a summary analysis.

Answer the five questions honestly, then read the summary.

Notice what happens. You are the topic. The LLM has no idea why you are interested in this. It cannot cheat, because the answer is not in its training data; the answer is inside you. Every word of signal comes from you. The LLM provides the probing and the structure. You provide the content.

Read the summary when it arrives. Some of it will be obvious. But look for the part that surprises you: a connection you had not made between two of your own answers, a reason you had not put into words before, a motivation you had not named.

That is the interview inversion producing something neither you nor the LLM could have produced alone.

Exercise 2: Build Your Cognitive Profile

Set aside fifteen to twenty minutes. Use the most advanced model available to you, with extended thinking or reasoning turned on if the option exists. This exercise asks the LLM to probe the structure of how you think; give it every advantage.

Paste this prompt:

Interview me. Your goal is to build a profile of how I think, how I approach problems, what I value, and what kinds of reading or learning I enjoy and why. The output will be a document called my-profile.

Start with these three questions:

Q1: Tell me about a problem you solved recently that you are proud of. Q2: When you are stuck on something, what does getting unstuck feel like? Q3: If you were explaining a problem to a friend, how would you start?

After those three questions, keep probing based on what my answers reveal. Number your questions sequentially (Q4, Q5, Q6...) so we can track them.

Keep going until you have enough to write the profile. Do not rush to write it. When you think you are ready, ask me first if there is anything we missed.

Do not wrap up early. The goal is depth, not speed. The interesting material is behind the obvious answers, not in them.

Ask no more than 21 questions. When you write the profile, keep it to 800 words or less.

When the profile is done, read it. Look for the parts that surprise you: observations you would not have written about yourself, connections between things you had not noticed were connected.

Then take that profile and paste it into your LLM's preferences or custom instructions. You have just calibrated the instrument. Every session from here starts better than it would have without it.

Exercise 3: Find the Codec Boundary

Open the photos on your phone and pick one of a place that matters to you. A room from your childhood. A friend's kitchen. The view from a trip you took years ago that you still think about.

Open any LLM that can generate images. Describe the scene in as much detail as you can. Take your time with it. Try to capture not just what is in the photograph but why it still matters to you. Then ask the model to generate an image from your description.

Look at the generated image next to your photograph. The facts will mostly be there. Something will be missing all the same, and you will feel it immediately.

That gap is the codec boundary. Your description was a compression of something richer, and no matter how carefully you chose the words, the light in the room, the reason the photo still matters years later, none of it survived the trip through language. If it did not survive your description, it did not reach the LLM either. The model can only work with what arrives through the channel, and what arrives is always a fraction of what you meant.

This is not a problem better models will solve. It is a property of language itself. Understanding this boundary is what makes the practice work: you stop expecting the LLM to read your mind and start learning to use the channel deliberately.

Exercise 4: Your First Waypoint

This exercise builds on Exercise 2. If you have not built your cognitive profile yet, go back and do that first.

Open a new conversation with any LLM. Use the most advanced model available with extended thinking turned on. In your first message, paste the full output from your Exercise 2 profile, then directly below it paste the following prompt. Replace the bracketed section with something real: an idea you have been carrying around, a question you keep coming back to, a problem you have not been able to frame clearly enough to make progress on.

I have an idea [________________] I'd like to explore, it's stuck in my head and I want to get it out. Interview me. The output will be a markdown document "waypoint-001.md" and should describe enough for me to start new sessions in the future to continue my exploration with a large language model. Ask me no more than three questions per turn, and ask no more than twenty-one questions. The output should target about five hundred to one thousand words max. Do not rush, keep going until you think we've explored enough to express the idea.

The "001" in the filename is not decoration. It means there will be a 002. The waypoint captures enough context that you can hand it to a fresh conversation weeks or months from now and pick up where you left off. That is the practice.

Read the waypoint the LLM produced. Look for the parts that surprised you: connections you did not expect, phrasing that captures something you have been carrying but never said out loud. Notice the difference between this exercise and the profile you built in Exercise 2. The profile was about you. The waypoint is about an idea you care about, explored through you. Calibration first, then exploration.

Some people who try this will write things they have wanted to write for years. Some will discover that the creative capacity they had been told they lacked was there the whole time, waiting for a channel wide enough to carry it. LLMs do not provide the creativity. They provide the reach, the patience, and the non-judgmental open channel. What comes through the channel is yours.