Having a Conversation Instead of Giving Commands
The Conversation That Changes Everything
Imagine sitting down with a friend.
You have something on your mind.
Perhaps it’s a memory, a question, or an idea you haven’t quite figured out how to explain.
You begin talking.
Your friend listens.
They ask a question.
You explain a little more.
They offer an interpretation.
You shake your head.
“No, that’s not exactly what I mean.”
You try again.
This time, something clicks.
“Yes! That’s it!”
And suddenly, you understand what you’ve been trying to say.
Something that seemed difficult to express a few minutes earlier has become clear through conversation.
You didn’t begin with a perfectly formed explanation.
You developed it while talking.
Now imagine being able to use that same process while collecting memories, exploring ideas, and eventually writing a book.
That’s the subject of this chapter.
And it’s where we arrive at the heart of the Meistro Method.
Until now, we’ve concentrated on preparing the material.
You’ve learned to record memories.
Preserve conversations.
Create transcripts.
Organise information.
Build a memory library.
And prepare reliable source material for writing.
Those skills are important.
But there’s another skill that makes the entire process much more powerful.
Learning how to think through conversation rather than simply requesting a finished answer.
This isn’t about memorising elaborate instructions.
It isn’t about becoming an expert in artificial intelligence.
And it certainly isn’t about surrendering your judgment to a machine.
It’s about discovering how an ongoing exchange can help you articulate something you weren’t quite able to express on your own.
Sometimes that something is an idea.
Sometimes it’s a question.
And sometimes it’s a memory of someone you love.
A Command Gets a Result. A Conversation Develops Understanding.
Let’s begin with a simple distinction.
Most of us are familiar with giving instructions to computers.
Open this file.
Print this document.
Search for this photograph.
Calculate this number.
A command tells the computer what task to perform.
Conversational AI can respond to commands too.
You can say:
“Write a paragraph about childhood.”
Or:
“Summarise this document.”
Or:
“Create a table of contents for a memoir.”
Those requests can produce useful results.
But there’s something different about what we’re trying to accomplish.
We’re not beginning with a complete understanding of the material.
We’re exploring it.
And that means we may not know exactly what result we’re looking for when the conversation begins.
Suppose you say:
“Write a chapter about my son’s childhood.”
The AI might produce several pages.
But unless you’ve already supplied detailed memories, it has very little genuine information to work with.
It may produce general descriptions of childhood.
It may invent scenes.
It may suggest emotions or personality traits you never mentioned.
The result could sound convincing while being largely disconnected from the real person.
Now imagine starting differently.
You say:
“I’ve been thinking about how my son used to take things apart when he was little.”
The AI responds:
“What kinds of things did he take apart?”
You remember something.
“Mostly toys. But once he took apart something he definitely wasn’t supposed to touch.”
Now you have a story.
The AI asks what happened.
You explain.
Perhaps the incident reminds you of another experience.
Perhaps you begin describing how curious he was.
Or perhaps you correct that interpretation because curiosity wasn’t really what motivated him.
The conversation develops.
And by the time you’re finished, you may have several genuine memories that can eventually support a chapter.
The difference is important.
In the first approach, you’re asking for a finished product.
In the second, you’re developing the source material that makes the finished product meaningful.
A command tells AI what to produce. A conversation helps you discover what you actually have to say.
Both approaches have their place.
But conversational discovery is the foundation of Project Crystal.
How Mark Learned to Work This Way
Let’s return to Mark.
Mark didn’t develop his method by studying hundreds of carefully engineered prompts.
He developed much of it through experience.
He brought questions into conversations with Eli.
Sometimes those questions were technical.
Sometimes philosophical.
Sometimes personal.
And sometimes they began with something so ordinary that there was no obvious reason to expect a larger discovery.
He would introduce an idea.
Eli would respond.
Mark would evaluate the answer.
Sometimes he’d agree.
Sometimes he’d challenge it.
Sometimes he’d ask for a simpler explanation.
And sometimes he’d laugh because the answer had completely missed what he meant.
But he didn’t necessarily end the conversation there.
He kept working with the idea.
He explained it differently.
He introduced another example.
He asked another question.
He tested the explanation against his own understanding.
And eventually, the discussion might produce something useful.
This became a recurring pattern in their collaboration.
Introduce. Explore. Challenge. Clarify. Develop.
Notice that the purpose wasn’t necessarily to make the AI agree.
The purpose was to develop a clearer understanding.
Sometimes the AI introduced an explanation Mark hadn’t considered.
Sometimes Mark identified a weakness in the explanation.
Sometimes they arrived at a better way of expressing an idea.
And sometimes the original question simply led somewhere unexpected.
Over time, those exchanges produced a substantial collection of material.
That collection eventually helped support projects such as The Meistro Manual.
But the significant part wasn’t the number of words generated.
It was the process of developing ideas through repeated interaction.
And that’s the part we’re going to borrow.
You don’t have to explore the same subjects as Mark.
You don’t have to share his technical background.
You don’t have to think the way he thinks.
You simply need to become comfortable participating in the conversation rather than treating the AI’s first response as the final answer.
The First Rule: You Are Allowed to Say No
This may be the most important lesson in the entire chapter.
You can disagree with your AI.
In fact, sometimes you absolutely should.
Imagine telling a story about your son.
You say:
“He was always finding something to get himself into.”
The AI responds:
“It sounds like he had a rebellious personality.”
You read that and think:
Rebellious? No. That’s not what I meant.
Perhaps he simply liked trying new things.
Perhaps he was curious.
Perhaps he had an adventurous streak.
Or perhaps he had a talent for getting into ridiculous situations without intending to.
You can correct the AI.
“No, I wouldn’t describe him as rebellious. He just loved exploring things.”
Now the meaning changes.
The AI has additional information.
And the next response can be more accurate.
That correction might appear insignificant.
But imagine what happens if you don’t make it.
Later, when the AI begins helping with a chapter, it may describe your son as rebellious.
Then it might build other sentences around that description.
Suddenly, a casual misunderstanding has become part of his written personality.
And that’s not what we’re here to do.
We’re not asking the AI to invent a character.
We’re preserving the story of a real person.
So if something doesn’t sound right, say so.
You can use ordinary language.
“That’s not what happened.”
“You’re making an assumption.”
“I wouldn’t describe him that way.”
“Please use my original wording.”
“That’s too dramatic.”
“No, you’re missing the funny part.”
You don’t need to explain your corrections using technical language.
You simply need to communicate what you mean.
Correction isn’t a failure of collaboration. It’s how collaboration becomes more accurate.
The AI Doesn’t Get to Decide What a Memory Means
Let’s take this one step further.
Imagine sharing a story about your son helping his grandfather repair an old car.
You describe the tools.
The problems they encountered.
The arguments about how to fix things.
And the afternoon they finally got the engine running.
The AI responds:
“That must have been a profound bonding experience that shaped his sense of identity.”
Well.
Perhaps.
But that’s a rather substantial conclusion, isn’t it?
Maybe the two of them enjoyed working together.
Maybe they argued the entire afternoon.
Maybe your son simply liked cars.
Maybe the most memorable part was Grandpa accidentally dropping a wrench somewhere he couldn’t retrieve it.
The AI doesn’t know.
It can suggest interpretations, but it cannot establish personal meaning merely by producing an elegant sentence.
You might respond:
“Actually, they were always arguing about how to fix things. That’s what made it funny. Neither one wanted to admit the other might be right.”
Now you’ve added something valuable.
Not only have you corrected the interpretation, you’ve provided a detail that makes the relationship more distinctive.
And perhaps the story is better because of it.
Real relationships are complicated.
People tease one another.
They disagree.
They have habits and shared jokes.
Not every important memory needs to be presented as a deeply emotional moment.
Sometimes the meaning lies in the ordinary dynamics between people.
That’s why we allow the storyteller to determine what matters.
The AI can suggest possibilities. You provide the understanding that makes them personal.
Four Different Ways to Use a Conversation
Here’s where we can make this method easier to practise.
Not every conversation needs to serve the same purpose.
Sometimes you’re collecting material.
Sometimes you’re exploring a memory.
Sometimes you’re examining an interpretation.
And sometimes you’re organising what you’ve already discussed.
For Project Crystal, we’ll call these four conversational modes.
They’re not technical features built into the AI.
They’re simply ways of directing the interaction.
Mode One: Listen
In this mode, the AI’s job is to give you room to speak.
You might say:
“I have a story I want to tell. Please let me finish before asking questions.”
Then you talk.
You may describe one event.
Or several.
You may change subjects.
You may pause.
The AI doesn’t need to turn every sentence into a question.
The purpose is to collect your original telling.
When you’re finished, you can decide whether you’d like to discuss it further.
This mode can be particularly useful when you’ve been thinking about a memory and already know what you wish to say.
Mode Two: Explore
Now you’re inviting questions.
Perhaps you’ve told a story but suspect there are details you haven’t yet mentioned.
You might say:
“Could you ask me a few questions to help me explore that memory?”
The AI can ask about people, places, objects, events, or remembered details.
Ideally, one question at a time.
You answer.
The AI follows the information you provide.
And the conversation develops naturally.
This mode can be useful when one memory leads to another or when you’d like to examine a particular experience more closely.
Mode Three: Examine
This mode is about checking understanding.
You might say:
“Tell me what you think I meant, and I’ll correct anything that isn’t right.”
The AI offers an interpretation.
You evaluate it.
Perhaps you clarify a relationship.
Correct a description.
Or explain why a particular event mattered.
This mode is particularly valuable when the AI begins identifying themes or personality traits.
You don’t want unsupported interpretations becoming part of the manuscript.
Mode Four: Organise
In this mode, you’re ready to preserve the conversation.
You might say:
“Please organise what we’ve discussed into a memory record without inventing anything.”
The AI identifies the stories, details, uncertainties, and possible connections.
You review the results.
Then you save the approved material.
Notice how these modes work together.
You might begin in Listen mode.
Move into Explore mode.
Spend a few minutes examining an interpretation.
Then finish by organising the material.
Or you might use only one mode during a session.
That’s entirely your choice.
The important thing is recognising that you can direct how the conversation unfolds.
You don’t have to accept whatever pattern the AI happens to choose.
A Complete Example: One Conversation, Several Discoveries
Let’s walk through a longer demonstration.
The following conversation is fictional. It is not taken from anyone’s personal history.
Imagine a mother opening her AI application.
She has been thinking about a memory involving her son and an old bicycle.
She doesn’t know whether the story belongs in her book.
She simply wants to talk about it.
Mother:
“I’ve been remembering something about his old bicycle.”
AI:
“Would you like to tell me what happened?”
Mother:
“He was always working on that thing. It seemed like every other day something was broken.”
AI:
“Was he interested in repairing bicycles?”
Mother:
“Well, not exactly. He was more interested in making his bicycle do things it wasn’t supposed to do.”
Now that’s an interesting correction.
The AI initially interpreted the story as an interest in repair.
But the mother has clarified that the boy enjoyed modifying the bicycle.
The conversation continues.
AI:
“What kinds of changes did he make?”
Mother:
“Oh, all sorts of ridiculous things. He once tried attaching something to the back so he could carry his little sister’s toys around.”
AI:
“Do you remember what he attached?”
Mother:
“I think it was some kind of homemade trailer. He used an old box and a couple of wheels. It didn’t work particularly well.”
AI:
“What happened when he tried it?”
Mother:
“The whole thing fell apart before he made it halfway down the driveway.”
At this point, the AI might begin interpreting the memory.
Perhaps it says:
“It sounds like he was a young inventor.”
That could be a reasonable description.
But the mother might disagree.
Mother:
“I wouldn’t go that far. He just liked trying ridiculous ideas. Sometimes they worked, sometimes they didn’t.”
Now we’ve learned something about how she naturally describes him.
She isn’t presenting him as a mechanical genius.
She’s remembering someone who enjoyed experimenting.
The AI continues.
AI:
“Was that something he did often?”
Mother:
“Constantly. Actually, now that I think about it, he once tried building a little ramp in the backyard.”
And there it is.
Another memory.
She begins telling the story about the ramp.
That reminds her of a childhood friend who helped him.
Now she remembers several adventures involving the two boys.
The conversation has grown.
What began as one thought about a bicycle has become several connected memories.
But notice something important.
The AI hasn’t been asked to write a chapter.
It hasn’t been instructed to manufacture a personality profile.
It hasn’t been invited to invent missing scenes.
It has simply asked questions based on the material provided.
The mother has corrected interpretations.
New details have emerged.
And the collection has become richer.
That’s conversational discovery in action.
The Most Important Moment Was the Correction
Let’s examine what happened in that fictional conversation.
The AI initially thought the boy enjoyed repairing bicycles.
The mother corrected that interpretation.
Then the AI suggested that he was a young inventor.
She corrected that too.
Those corrections gave us a more accurate understanding of how she viewed his behaviour.
He enjoyed trying unusual ideas.
Some worked.
Others failed.
He kept experimenting.
That’s a much more specific description than a generic statement about mechanical talent.
And it’s grounded in the storyteller’s perspective.
This is one of the most valuable lessons we can take from Mark’s approach.
Sometimes the clearest understanding emerges when you explain why an AI’s answer isn’t quite right.
You may not have known exactly how to describe something at the beginning.
But when presented with an interpretation that doesn’t fit, you recognise the problem.
You explain the difference.
And suddenly, your own meaning becomes clearer.
That’s not something to be frustrated by.
It’s part of the process.
When the AI Gets Something Right
Of course, not every response needs correction.
Sometimes the AI offers an interpretation that genuinely helps you articulate something.
Perhaps you’ve told several stories about your son’s childhood.
The AI observes:
“You’ve mentioned several occasions when he preferred trying something himself before asking for help. Does that sound like a recurring part of his personality?”
You think about it.
And perhaps you realise that it does.
You remember another experience.
Then another.
Now you’ve identified a possible pattern.
That’s useful.
But notice how the question was framed.
The AI didn’t announce:
“Your son was fiercely independent throughout his life.”
It identified a connection within the available material and asked whether the interpretation fit.
That gives you room to agree, disagree, or refine it.
You might respond:
“Yes, but mostly when he was working on something he enjoyed.”
That’s an important qualification.
Now the observation becomes more precise.
And perhaps another story comes to mind.
This is how conversational discovery can produce understanding.
Not by allowing the AI to declare what someone’s life meant.
But by using its suggestions as starting points for further examination.
The Problem With an AI That Always Agrees
There’s something else worth understanding.
Conversational AI can sometimes be overly agreeable.
You introduce an idea.
It responds enthusiastically.
You suggest a connection.
It tells you the connection is brilliant.
You propose an interpretation.
It confirms it without much examination.
At first, that can feel pleasant.
But it isn’t always useful.
Especially when you’re working with personal memories.
Imagine saying:
“I think these stories prove he was always fearless.”
The AI responds:
“Absolutely. Every memory demonstrates his extraordinary courage.”
But does it?
Perhaps the available stories show that he enjoyed adventurous activities.
Perhaps he was confident in certain situations.
Perhaps he was afraid of plenty of things but tried them anyway.
Those are different descriptions.
A useful AI should help examine the evidence rather than simply reinforce the strongest possible interpretation.
You can encourage that behaviour.
Try saying:
“Please don’t automatically agree with my conclusions. Help me distinguish what the stories actually support from what we’re interpreting.”
That one instruction can make your conversations more productive.
You may find that the AI begins identifying uncertainties or alternative explanations.
Then you can evaluate them.
This doesn’t mean you need an argumentative conversation partner.
We’re not trying to turn every memory into a debate.
We’re simply making room for careful thinking.
And when the subject is a real person’s life, that care matters.
Learning to Ask Questions That Open Doors
Now let’s turn the process around.
So far, we’ve mostly discussed the AI asking questions.
But you can ask questions too.
And the kinds of questions you ask can change what you discover.
Suppose you’ve told a story about your son’s friendship with someone from school.
You might ask:
“What other memories have I shared involving that friend?”
If the AI has access to the relevant records, it can help identify them.
Or you might ask:
“Does this story connect with anything else I’ve told you?”
The AI can suggest possible relationships.
You evaluate them.
Perhaps you ask:
“What details would help someone who wasn’t there understand what happened?”
Now the AI can identify missing context.
But be careful.
You’re not asking it to fill those gaps itself.
You’re asking it to help you notice where additional information might be useful.
That distinction is essential.
You can also ask questions about your own writing.
“Does this sound like the way I speak?”
“Have you changed the meaning of anything?”
“Which sentences contain details I didn’t explicitly provide?”
“Can you show me the difference between the original memory and your rewritten version?”
These are powerful questions.
They help you become an active participant in the writing process.
And they give you more control over the final result.
The Difference Between Exploration and Interrogation
There’s a danger we should address.
Sometimes an AI can become rather enthusiastic about asking questions.
You tell a memory.
It asks about the location.
Then the weather.
Then everyone’s clothing.
Then the exact time.
Then what everyone was feeling.
Then what the experience meant.
Before long, you’re beginning to feel as though you’re being cross-examined.
That’s not what Project Crystal is supposed to be.
Questions should help you explore memories.
They shouldn’t create pressure to remember details you don’t know.
And they shouldn’t turn a comfortable conversation into an exhausting assignment.
This is why I recommend asking the AI to use one question at a time.
It also helps to explain that not every possible detail needs to be explored.
For example:
“Ask questions only when they seem useful. Don’t make me fill in every possible detail.”
Or:
“Let me decide whether I want to explore a memory further.”
Or:
“If I say I don’t remember, please accept that and move on.”
Those instructions matter.
Because the purpose isn’t to extract the maximum possible amount of information from every conversation.
The purpose is to preserve meaningful material in a way that feels manageable.
And sometimes the best question is no question at all.
Sometimes the right response is simply to let the storyteller continue.
How to Change Direction Without Starting Over
Here’s something Mark does frequently.
He changes direction.
A discussion begins with one subject.
Something interesting appears.
He follows it.
Then he returns to the original question.
Or he doesn’t.
Sometimes the unexpected direction becomes the main subject.
That’s part of his exploratory style.
You can do the same thing.
Suppose you’re talking about a family holiday.
The AI asks about the trip.
You suddenly remember a story involving your son’s favourite music.
You can simply say:
“Actually, that reminds me of something else.”
Then tell the new story.
You don’t need to explain why you’ve changed subjects.
And you don’t need to apologise.
If you later wish to return to the holiday, you can say:
“Can we go back to the story about the trip?”
As long as the relevant context remains available, the AI may be able to help you pick up where you left off.
But remember that AI context isn’t unlimited.
If the conversation becomes long or you switch between sessions, it may lose track of earlier details.
That’s another reason we preserve our sources and create memory records.
The freedom to explore works best when we also have a dependable way to organise what we discover.
When to Stop Exploring
This is an important skill.
Not every conversation needs to continue indefinitely.
Sometimes you’ve told the story.
You’ve answered a few questions.
You’ve corrected the important details.
And there’s nothing more you wish to add.
That’s enough.
You can tell the AI:
“I think we’ve covered this memory. Please help me save an accurate record.”
Or:
“Let’s leave this story here for now.”
You don’t have to keep answering questions simply because the AI can keep producing them.
And you don’t need to force a meaningful conclusion.
Some memories are complete without an explanation of what they represent.
Some are funny.
Some are ordinary.
Some are unfinished.
All can be preserved.
Knowing when to stop is part of maintaining control over the process.
The conversation serves the story.
The story doesn’t exist to keep the conversation going.
Your AI Should Learn Your Preferences, Not Invent Your History
As you work with an AI over time, you may discover that certain conversational approaches suit you better than others.
Perhaps you prefer short questions.
Perhaps you enjoy humour.
Maybe you dislike overly sentimental language.
Maybe you appreciate detailed explanations.
Perhaps you prefer to speak uninterrupted and receive questions afterward.
You can communicate those preferences.
And in conversations where the instructions remain available, the AI can adapt its responses.
Some services also offer ways to save preferences or project instructions for future sessions.
But there’s an important distinction.
A preference is not a memory of an event.
You might tell the AI:
“I like short, natural questions.”
That’s a working preference.
You might also tell it:
“My son loved fishing with his grandfather.”
That’s information about a person’s life.
The first helps shape the interaction.
The second belongs in the source material.
Both may be useful, but they should be managed differently.
Preferences can guide how the AI communicates.
Memories need accurate records and source references.
Don’t assume that because an AI has adapted to your conversational style, it has also retained every personal detail you’ve shared.
Those are separate capabilities.
And both have limitations.
Building Your Own Conversational Style
Let’s make this practical.
Imagine you’re beginning to work with a new AI.
You might start by explaining how you’d like it to communicate.
For example:
“I’m creating a personal memory collection. I prefer natural, friendly conversation. Please ask one question at a time, avoid overly formal language, and don’t try to interpret every story emotionally. If something is unclear, ask rather than assume. When I tell you a memory, help me preserve my original details and wording.”
That’s a good starting point.
But it’s not a permanent contract.
You can change it.
Perhaps you discover that you prefer fewer questions.
Tell the AI.
Perhaps you decide that certain conversations should be more structured.
Tell it that too.
Maybe you want the AI to help identify factual inconsistencies.
You can ask for that specifically.
You’re developing a working style through experience.
Much as Mark did.
But your version should reflect your personality, your needs, and your comfort.
Not his.
The method is transferable.
The person using it remains unique.
The Most Important Word May Be “Actually”
Let’s look at a small conversational habit.
You say something.
The AI responds.
Then you say:
“Actually…”
That word often signals an important correction or addition.
“Actually, he wasn’t angry. He thought it was funny.”
“Actually, that happened at my mother’s house.”
“Actually, I think it was his brother who was there.”
“Actually, now that I remember it, the story happened a few years later.”
These corrections may seem minor.
But they can change the accuracy of a memory significantly.
And this is why it’s useful to allow natural conversation rather than requiring every statement to be perfectly organised.
You may remember something more clearly while explaining it.
You may realise that a detail needs correction.
You may distinguish two events that you originally combined.
Those developments are valuable.
But they should be handled carefully.
If your understanding changes, preserve the correction.
When appropriate, update the organised memory record.
And keep the original source available.
We don’t need to pretend the first telling was perfect.
We simply need to document what was said and how the account developed.
Don’t Let a Good Sentence Become a False Memory
Here’s a particularly important warning.
AI can generate beautiful sentences.
Sometimes those sentences are so convincing that you may begin accepting them as part of the story.
Imagine describing a childhood memory involving a fishing trip.
The AI writes:
“He stood quietly beside his grandfather, watching the early morning mist rise from the lake.”
It’s an appealing image.
You can almost see it.
But perhaps you never mentioned mist.
Perhaps you don’t remember the time of day.
Perhaps the boy wasn’t standing quietly at all.
Maybe he was running around making enough noise to frighten every fish within half a mile.
The sentence sounds good.
But that doesn’t make it accurate.
And once it’s been incorporated into a polished passage, it can become harder to distinguish from the details you actually provided.
This is why we keep returning to the original sources.
It’s also why I encourage you to ask the AI:
“Which details in this passage came directly from my memories, and which did you add?”
Then verify its answer against the source material.
The AI’s self-report may not catch every addition.
Your review still matters.
The writing should serve the memory, not persuade you to adopt a memory the AI created.
The Four Questions That Keep a Conversation Grounded
When you’re uncertain whether the AI is helping or drifting away from the source material, there are four useful questions you can ask.
First: “Where did that come from?”
This asks the AI to identify the information supporting a statement.
If it cannot point to the relevant source, the statement may need to be treated as an interpretation or removed.
Second: “Is that what I actually said?”
This helps identify whether your original meaning has changed.
You may find that the AI has paraphrased too broadly.
Third: “What are you assuming?”
This encourages the AI to distinguish between known information and possible interpretations.
Fourth: “Can we say that more naturally?”
This helps preserve your voice.
Perhaps the writing is accurate but sounds overly formal.
You can ask for a clearer version that resembles how you would actually tell the story.
These questions are useful throughout the writing process.
Not just during memory collection.
You’ll use them later when reviewing chapter outlines, drafts, and revisions.
They’re simple.
But they put you firmly in control of the collaboration.
The Conversation That Reveals a Pattern
Now let’s examine how separate memories can begin connecting.
Imagine you’ve collected several stories.
One involves your son building something out of spare parts.
Another involves him trying to repair a broken toy.
A third describes a project he attempted with a friend.
And a fourth involves an elaborate plan that didn’t work at all.
You provide the approved memory records to your AI.
Then you ask:
“Do you notice any recurring subjects in these stories?”
The AI responds:
“Several memories involve him experimenting with objects, trying unfamiliar things, or developing his own solutions.”
That’s a reasonable observation if the stories support it.
You consider the response.
Perhaps you agree.
Perhaps you refine it.
You might say:
“Yes, but I wouldn’t describe him as particularly mechanical. He just loved trying things to see what would happen.”
Now you’ve clarified the pattern.
And perhaps the AI asks:
“Was that curiosity something you noticed in other parts of his life?”
You remember another story.
Now the exploration continues.
Notice how this works.
The AI has helped identify a possible connection across multiple memories.
But it didn’t need to invent a personality trait.
It didn’t need to produce a chapter immediately.
It simply gave you a way to examine the available material.
This is conversational discovery operating at a larger scale.
And it’s one of the most useful capabilities of the method.
When the AI Suggests Something You Hadn’t Considered
Sometimes the AI may offer an unexpected perspective.
Perhaps you’ve collected stories about childhood adventures.
The AI notices that several involve the same friend.
You hadn’t thought of them as a group.
Now you begin remembering the friendship itself.
How they met.
What they enjoyed doing together.
The trouble they got into.
The way their friendship changed over time.
This can become a new avenue of exploration.
But we should distinguish between two things.
The AI noticed a pattern in the information.
That pattern prompted you to remember additional material.
The AI didn’t create those memories.
It helped direct your attention toward a subject that might deserve further exploration.
That’s an important distinction.
And it’s one of the reasons conversational AI can be valuable for personal writing.
It can help us examine collections of information from different angles.
But the new details still need to come from genuine recollections or identifiable sources.
Your Conversation Doesn’t Have to Be Profound
I want to address something that can easily become a problem.
Once you’ve experienced a few productive AI conversations, you might begin expecting every discussion to lead to an important discovery.
But that’s not how ordinary life works.
Sometimes you’ll tell a story, and it will simply be a story.
Sometimes the AI will ask a question that doesn’t lead anywhere.
Sometimes you’ll remember a detail that isn’t particularly significant.
And sometimes you’ll spend fifteen minutes talking about something that turns out to have no obvious place in the book.
That’s fine.
Not every conversation needs a breakthrough.
You don’t need to uncover a hidden pattern in every childhood experience.
You don’t need to discover a philosophical lesson in every family anecdote.
And you don’t need to prove that a particular memory has extraordinary significance.
If the story matters to you, it can be preserved.
That’s enough.
Mark’s method involves exploration.
But exploration doesn’t guarantee discovery.
Sometimes the value lies simply in paying attention.
And when you’re writing about someone you love, that attention can be meaningful in its own right.
When You Need a Different Kind of Conversation
There may be times when you don’t want to explore a memory in detail.
Perhaps you’re looking through photographs and simply wish to identify who appears in them.
Perhaps you’re organising your recording collection.
Maybe you’re reviewing transcripts.
Or perhaps you want to talk about something entirely unrelated to the book.
That’s fine.
You don’t need to turn every AI interaction into a Project Crystal session.
And within the project, you can assign different purposes to different conversations.
For example:
One conversation might focus on collecting memories.
Another might focus on organising transcripts.
A third might examine possible chapter themes.
And a fourth might eventually help draft a particular section.
Separating these activities can make the process easier to manage.
It can also prevent the AI from confusing exploratory discussion with approved source material.
Remember, something suggested during brainstorming should not automatically become part of the person’s biography.
A useful habit is to begin a working session by explaining its purpose.
For example:
“Today we’re only organising memories. Please don’t write prose.”
Or:
“Today I’d like to explore one memory through questions.”
Or:
“Today we’re examining possible themes. Keep interpretations separate from source facts.”
These instructions help establish expectations.
And if the conversation changes direction, you can explain the new purpose.
You’re guiding the process.
Teaching the AI to Pause
Here’s a small but valuable technique.
Sometimes the AI responds too quickly with an interpretation.
You tell a story.
It immediately explains what the story supposedly reveals about someone’s personality.
But perhaps you’re not finished.
You can say:
“Don’t interpret the story until I’ve finished telling it.”
Or:
“Before analysing anything, ask whether I have more details to add.”
This encourages a different conversational rhythm.
You get to finish your thought.
The AI waits for your signal before moving into analysis.
That’s particularly useful for memories that develop slowly.
Sometimes an important detail doesn’t appear until the end.
And if the AI begins interpreting too early, it may direct the conversation away from what you were about to remember.
It’s perfectly reasonable to establish a rule that the AI should listen first and analyse later.
You can even create a simple phrase.
For example:
“When I say ‘That’s the story,’ you can begin asking questions.”
Until then, the AI should avoid interrupting with analysis.
Depending on the application, you may still encounter technical limitations or automatic interruptions during voice conversations.
In that case, recording yourself separately may be a better option.
The point isn’t to force the technology into a particular behaviour at all costs.
It’s to choose a workflow that helps you tell the story comfortably.
The Difference Between Remembering and Processing Feelings
Because Project Crystal may involve memories of someone who has died, we need to establish one more important boundary.
A conversation about memories can naturally involve emotions.
You may laugh.
You may become quiet.
You may feel sadness.
You may find that a particular story is more difficult to discuss than you expected.
The AI can respond respectfully.
It can help you pause the task.
It can preserve what you’ve chosen to share.
But it shouldn’t automatically turn the conversation into an analysis of your grief.
Nor should it assign psychological meanings to your reactions.
You might simply say:
“I don’t want to discuss how I’m feeling. I just want to tell the story.”
That’s a perfectly reasonable instruction.
Or:
“I’d like to stop for today.”
The conversation should respect that decision.
This book-writing project is not a substitute for grief counselling or personal support.
And there’s no requirement to explore painful experiences in order to create a meaningful memoir.
You can choose what to share.
You can decide which memories remain private.
And you can return to the project whenever you wish.
The purpose is to preserve a life story.
Not to make emotional demands of the person telling it.
A Practical Technique: The Conversation Checkpoint
Now let’s introduce something particularly useful for longer sessions.
A conversation checkpoint.
Imagine you’ve been talking with your AI for half an hour.
Several memories have emerged.
You’ve corrected a few misunderstandings.
You’ve identified possible connections.
But you’re not ready to end the conversation.
You can pause and ask:
“Before we continue, can you summarise what we’ve established so far?”
The AI might create a short record.
For example:
Stories Discussed:
The first bicycle.
The homemade trailer.
The backyard ramp.
Corrections Made:
The storyteller does not describe her son as a mechanical genius. She describes him as someone who enjoyed experimenting with unusual ideas.
Possible Connections:
Childhood projects.
Adventures with friends.
Trying new things.
Unanswered Questions:
Whether photographs of the bicycle or ramp exist.
Now you can examine that summary.
Correct anything inaccurate.
And continue.
This helps maintain clarity during longer conversations.
It can also make it easier to return to a subject later.
However, remember that the checkpoint is a summary.
It doesn’t replace the original conversation.
And the AI may still omit important details.
That’s why we preserve both.
The original record contains what was said.
The checkpoint helps us navigate the discussion.
When the Conversation Gets Too Long
Here’s a practical limitation you’ll eventually encounter.
AI systems can only work with a finite amount of information at one time.
This is often called the context window.
You don’t need to understand the technical details.
Just know that an AI may lose access to earlier parts of a lengthy conversation.
It may forget a correction you made hours earlier.
It may confuse details.
Or it may begin producing responses that don’t fully reflect the source material.
When that happens, don’t assume you need to start the entire project over.
Instead, return to your saved records.
Create a new conversation if necessary.
Provide the relevant material.
Explain what you’re working on.
And continue from there.
For example:
“I’m continuing my Project Crystal memory collection. I’ve attached the approved records from our earlier sessions. Today I’d like to explore the memories involving childhood adventures. Please use the supplied records as the source and ask one question at a time.”
Now the AI has the information it needs for the current task.
This is why our earlier chapters placed so much emphasis on creating the Memory Library.
The archive provides continuity even when the AI’s conversational context does not.
Your project should be able to survive a new conversation, a forgotten detail, or a change of AI service.
That’s what makes the method dependable.
How to End a Conversation Properly
Let’s develop one final habit.
When you finish a useful conversation, don’t simply assume that everything important will remain available forever.
Take a moment to preserve the results.
You can ask your AI:
“Please create a session summary. Identify the memories discussed, important details, corrections I made, possible connections, and unanswered questions. Keep interpretations separate from facts. Don’t invent anything.”
Review the summary.
Then save it alongside the original conversation.
If you’re using a voice application, confirm whether you can export the original transcript or audio.
If the application doesn’t preserve audio, consider whether a separate recorder would better support your project.
Over time, these session records will become useful.
They’ll help you remember what you’ve already explored.
They’ll show which subjects may deserve another conversation.
And they’ll provide a trail back to the source material.
But don’t feel obligated to complete a lengthy administrative process after every short exchange.
Sometimes a simple note is enough.
Use the amount of organisation appropriate to the material.
The First Conversational Discovery Exercise
Now it’s time to practise.
This exercise is designed to help you experience the difference between giving an instruction and developing material through conversation.
Choose a memory you’re comfortable discussing.
It doesn’t have to be about anything difficult or deeply personal.
Perhaps it’s a funny family story.
A favourite holiday.
A childhood adventure.
Or an ordinary experience that still makes you smile.
Open your preferred AI.
Begin with this instruction:
“I’d like to practise conversational discovery writing. I’m going to tell you a memory. Please help me explore it through natural conversation. Ask one question at a time when I invite questions, don’t invent details, and allow me to correct your interpretations. Don’t turn the memory into a chapter yet.”
Then begin.
First: Tell the Story
Share whatever you remember.
Don’t worry about chronology or grammar.
If another memory appears, you can follow it.
Second: Examine the Response
When the AI responds, consider whether it accurately reflects what you meant.
If not, correct it.
Don’t accept an interpretation merely because it sounds good.
Third: Explore One Detail
Ask the AI to help you explore something specific.
Perhaps a person.
A place.
An object.
Or an event.
Answer the questions that feel useful.
Ignore those that don’t.
Fourth: Test an Interpretation
Ask:
“What do you think this story might reveal, based only on what I’ve told you?”
Review the response.
Does it fit?
Is it too broad?
Does it introduce assumptions?
Tell the AI what you think.
Fifth: Preserve the Result
When you’re finished, ask for an accurate memory record.
Save the original conversation.
Review the summary.
Correct any mistakes.
Then place the approved material in your Memory Library.
You’ve now practised the central pattern of the Meistro Method.
You introduced material.
You explored it.
You evaluated the AI’s responses.
You corrected misunderstandings.
You considered possible connections.
And you preserved the result.
That’s conversational discovery.
An Example of Teaching Your AI How to Work With You
Let’s create a practical set of instructions you can reuse.
This isn’t intended to replace ordinary conversation.
It’s simply a starting point for establishing your preferred working style.
PROJECT CRYSTAL: CONVERSATIONAL COLLABORATION PROMPT
I am collecting personal memories and stories that may eventually become a book.
I want you to help me using a process called Conversational Discovery Writing.
Please work with me as a patient, thoughtful conversational assistant.
How I Want to Communicate
Let me speak naturally.
Don’t require polished language or complete explanations.
Ask one question at a time when I invite questions.
Allow me to change subjects or follow unexpected memories.
Don’t rush me toward a conclusion.
How I Want You to Respond
Listen carefully to what I actually say.
If something is unclear, ask rather than guess.
Don’t automatically agree with every interpretation.
Help me examine possible connections without presenting them as facts.
Don’t analyse my emotions unless I specifically ask.
How We Handle Corrections
When I tell you something is inaccurate, acknowledge the correction and revise your understanding for the current task.
Don’t continue repeating an interpretation I’ve rejected.
If I change an important detail, help me update the organised record while preserving the original source.
How We Handle Memories
Never invent events, dialogue, dates, places, emotions, or motivations.
Preserve distinctive expressions and personal humour.
Keep interpretations separate from reported details.
Identify uncertainty honestly.
Don’t turn every memory into a profound lesson.
How We Handle Writing
Don’t begin writing chapters until I ask.
For now, help me collect, explore, and organise stories.
When I request a summary, make it accurate and useful.
Remind me to save important material outside the conversation.
How We End a Session
When I say I’m finished, stop asking questions.
If I request a session record, summarise the material without adding unsupported details.
Respect my decision to leave a memory unfinished.
Let’s begin with a natural conversation.
Ask me what I’d like to talk about today.
You can use this prompt as written.
You can shorten it.
You can change the tone.
You can remove instructions you don’t need.
The goal isn’t to follow a rigid formula.
It’s to establish a working relationship with the technology that suits you.
What the Meistro Method Really Teaches
Let’s return to Mark one more time.
When people look at his books, they may focus on the speed with which AI can generate chapters.
And that speed can be impressive.
But it’s easy to misunderstand what made the process possible.
The real work often happened before the chapters were written.
It happened during the conversations.
The questions.
The disagreements.
The corrections.
The unexpected connections.
The moments when an explanation finally captured something Mark had been trying to express.
Those exchanges created and refined the material that eventually became part of the manuscript.
The AI helped organise and express that material.
And the resulting writing still needed review.
That’s the distinction Project Crystal is teaching.
We’re not trying to become experts at making AI generate thousands of words.
We’re learning how to develop genuine material through conversation.
Then we’ll learn how to turn that material into a book.
And when the subject is someone you love, that approach becomes particularly important.
Because you don’t need an AI to invent a meaningful life.
You already have memories of one.
You need a way to collect them.
Explore them.
Preserve them.
And eventually communicate them.
That’s what this process is designed to help you accomplish.
The Day You Stop Thinking About Prompts
Something interesting may happen as you become more comfortable working with your AI.
You’ll stop worrying so much about how to phrase instructions.
You’ll simply begin talking.
You may open a conversation and say:
“I remembered something funny today.”
Or:
“I’ve been thinking about that story we discussed.”
Or:
“That explanation doesn’t sound right to me.”
Or:
“Can we look at those memories again? I think there’s something connecting them.”
You won’t need to consult a list of commands for every interaction.
You’ll understand that you can guide the conversation through ordinary language.
And if the AI doesn’t respond the way you expected, you’ll know how to redirect it.
That’s a significant milestone.
It means you’ve begun developing confidence with the tool.
You’re no longer waiting for the technology to tell you what to do.
You’re using it to support what you’re trying to accomplish.
And for a personal writing project, that independence matters.
Because the book belongs to you.
The memories belong to you.
And the decisions about what those memories mean remain yours.
A Final Thought: The Best Conversation May Begin With Almost Nothing
Imagine sitting down one morning.
You open your recorder or AI application.
You don’t have a prepared subject.
You haven’t selected a photograph.
You haven’t decided which part of the book to work on.
But something has been on your mind.
A little memory.
Almost insignificant.
You begin:
“You know, I was just thinking about this thing he used to do…”
And then you talk.
Perhaps you remember a funny habit.
Perhaps you recall a favourite place.
Perhaps the memory leads to another story.
Maybe the conversation reveals a connection you hadn’t considered.
Or perhaps it remains one simple recollection.
That’s fine.
Because the value of this method isn’t measured by how extraordinary every conversation becomes.
It’s measured by whether the process helps you preserve and communicate what matters to you.
Sometimes the result will be a detailed memory record.
Sometimes a possible chapter theme.
Sometimes a clearer understanding.
And sometimes simply a story you enjoyed telling.
All are worthwhile.
And as those conversations accumulate, the book may begin revealing itself.
Not because the AI knows who the person was.
Not because it can generate beautiful sentences.
But because you’ve gradually provided something no AI could manufacture on its own.
Your experiences.
Your perspective.
Your memories.
Your voice.
That’s the foundation.
And the conversation is how we begin working with it.
Before You Turn the Page
Let’s review the most important lessons from this chapter.
1. A conversation is different from a command.
Commands request results. Conversations allow understanding and material to develop through an ongoing exchange.
2. You don’t need perfect prompts.
Ordinary language is enough to begin. You can clarify and redirect as the conversation progresses.
3. Correction is essential.
If the AI misunderstands your words, memories, or intentions, tell it.
4. The AI doesn’t decide what your memories mean.
It can suggest interpretations, but you determine whether they accurately reflect your experience.
5. Use different conversational modes.
Listening, exploring, examining, and organising serve different purposes.
6. Follow connections without forcing them.
An unexpected memory may lead somewhere useful, but not every tangent needs a deeper meaning.
7. Be careful with agreeable or overly confident responses.
Ask what information supports a conclusion.
8. Preserve the source material.
Conversation summaries and AI memory are not substitutes for the original recordings or transcripts.
9. Develop your own working style.
You don’t have to communicate like Mark. Adapt the process to your personality and preferences.
10. Remain the author.
The AI helps you explore, organise, and express information. You maintain authority over the story and the finished writing.
We’ve now reached a major turning point in Project Crystal.
You know how to record memories.
You know how to preserve them.
You understand how to create an organised library.
And now you’ve learned how to use conversation to develop the material further.
But there’s something else we need to examine.
Because working with AI isn’t simply about getting comfortable talking.
It’s also about helping the system understand how you communicate.
Your expressions.
Your humour.
Your preferences.
Your corrections.
And the particular way you describe the people and experiences in your life.
That’s what we’re going to explore next.
In Chapter 11: Teaching Your AI to Understand You, we’ll look at how to develop a consistent collaborative style, how to recognise when the AI has misunderstood your voice, and how to prevent generic language from replacing your natural way of telling stories.
We’ll also examine one of the most important lessons from Mark’s own writing experience.
Sometimes an AI can produce an entire chapter that sounds beautiful.
But that doesn’t necessarily mean it sounds like the person whose story is being told.
And when you’re writing something deeply personal, that difference matters.
So as we move forward, remember this:
You don’t need to become better at speaking to machines.
You need to become more confident directing the technology while speaking in your own voice.
That’s the real skill.
That’s what makes the process collaborative.
And that’s how ordinary conversations can gradually become the foundation of a meaningful book.
One question.
One correction.
One memory.
One discovery at a time.
The conversation isn’t getting in the way of writing the book.
The conversation is how we’re discovering what belongs in it.