Meet Your New Writing Partner

So, What Exactly Are We Working With?

Let’s get one thing out of the way immediately.

You do not need to understand how artificial intelligence works internally to use it effectively.

You don’t need to understand computer programming.

You don’t need to know what a neural network is.

And you certainly don’t need to learn an entirely new technical vocabulary just to tell a story about your family.

Think about the first time you used a smartphone.

You probably didn’t study radio frequencies, computer processors, or the engineering behind touchscreen displays.

You learned where to touch the screen, what the applications did, and how to accomplish whatever you needed.

Artificial intelligence can be approached in much the same way.

You learn what it’s capable of, what its limitations are, and how to use it for the things that matter to you.

But there’s an important difference between using most computer applications and using conversational AI.

Traditional software generally expects you to learn its commands.

You click a button, select an option, and the software performs a particular function.

Conversational AI allows you to describe what you’re trying to accomplish in ordinary language.

You can explain your goal.

Ask questions.

Change your mind.

Correct a misunderstanding.

Request another explanation.

And continue the discussion without having to start the entire process over.

That’s what makes it particularly useful for our kind of writing.

You don’t have to learn to speak the computer’s language. You can begin by speaking your own.

And that changes what becomes possible.

The Machine That Can Hold a Conversation

Modern conversational AI systems are trained using enormous quantities of information.

Through that training, they develop the ability to recognise and generate patterns in language, follow instructions, answer questions, summarise information, and produce many different kinds of written material.

The systems behind these abilities are commonly called large language models.

You may hear the abbreviation LLM.

Don’t let the terminology intimidate you.

For our purposes, what matters is that an LLM can process the words you provide and generate a response that relates to your conversation.

You can ask it to explain something.

You can give it a passage to examine.

You can describe an experience and ask it to help you organise the details.

You can share a transcript and request a summary.

And you can ask it to transform an approved collection of memories into a draft chapter.

However, there is something important to understand.

An AI can produce convincing language without actually knowing whether every statement it makes is true.

Its ability to generate a fluent paragraph is not proof that the paragraph is accurate.

That distinction is especially important when working with personal memories.

Suppose you tell an AI about a birthday party from years ago.

You mention a cake, several relatives, and a funny incident involving the family dog.

The AI might generate a lovely paragraph describing the event.

But if it adds colourful decorations, a sunny afternoon, or a particular conversation that you never mentioned, it has crossed a line.

It has invented details.

Those details might make the writing sound more complete, but they don’t make the memory more accurate.

And our goal is not to manufacture a believable life story.

Our goal is to preserve a real one.

We’ll return to this principle throughout the guide because it is one of the most important safeguards in the entire process.

A Writing Partner, Not a Witness

I want to introduce a distinction that will help you understand how to work with AI.

Imagine you’re telling a friend about something that happened many years ago.

Your friend listens, asks questions, and helps you remember details.

They may notice something interesting in your story or ask whether two events were connected.

But unless they were there, they cannot independently confirm what happened.

They are working from what you tell them.

Conversational AI operates with a similar limitation when dealing with your personal experiences.

It can help you explore a memory.

It can identify important details.

It can point out that you’ve mentioned the same place in several different stories.

It can help organise events into a possible timeline.

But it wasn’t present for those experiences.

It doesn’t possess your memories.

And it doesn’t know the person you’re writing about simply because you have described them.

This is why I prefer to think of AI as a writing partner rather than a witness.

The phrase “writing partner” is a useful description of its role, not a claim that the software is a human being.

It provides conversational assistance.

You provide the lived experience.

You remain responsible for deciding whether the resulting material is accurate and appropriate.

That relationship can become remarkably productive.

But only when you understand which participant is responsible for what.

Choosing an AI That Works for You

There are several conversational AI services available.

Three familiar examples are ChatGPT, Claude, and Gemini.

They are developed by different companies and have different interfaces, features, limitations, and subscription options.

You may already use one of them.

If you do, that’s probably the best place to begin.

You don’t need to change applications simply because someone else prefers a different one.

The goal of Project Crystal is not to convince you to use one particular company’s software.

It’s to teach you a method that can work across different conversational systems.

ChatGPT

ChatGPT is developed by OpenAI.

You can interact with it through written conversations, and supported versions offer voice interactions and file-handling capabilities.

You can ask questions, share writing, organise information, and work through ideas over multiple exchanges.

Mark uses ChatGPT extensively in his own creative work.

His assistant is called Eli.

Over time, their conversations became a workspace for investigating ideas, developing writing projects, and creating The Meistro Manual.

But you don’t need Mark’s history or experience to begin using it.

You can start with a new conversation and a single story.

Claude

Claude is developed by Anthropic.

It is another conversational AI that can assist with writing, analysis, document review, and organising ideas.

Depending on the version and features available to you, it can be useful for working with lengthy documents or examining portions of a manuscript.

Like any AI, it can misunderstand instructions or produce inaccurate material.

Its writing still needs your review.

Gemini

Gemini is developed by Google.

It offers conversational assistance and, depending on the product and plan, can work with different kinds of information and files.

People who already use Google’s applications may find aspects of its interface familiar.

As with the other services, the important thing is learning how to communicate what you need and how to verify the results.

Which One Is Best?

For this project, the answer is surprisingly straightforward.

The best starting point is usually the AI you’re most comfortable talking to.

You don’t need to select a permanent favourite.

You don’t need to compare every feature.

And you certainly don’t need to subscribe to three different services before recording your first memory.

The features, costs, usage limits, and privacy options of these services change over time.

Before paying for anything, check the provider’s current information.

For now, choose a conversational AI you’re comfortable exploring.

We’ll focus on the skills that transfer from one service to another.

Do You Need a Paid Subscription?

Not necessarily.

Many AI services offer free access with certain limitations.

Paid plans may provide additional usage, access to different models, larger file allowances, or more extensive tools.

Exactly what is included depends on the service and can change.

Here’s the important part:

You don’t need advanced features to learn conversational writing.

A basic text conversation is enough to practise.

You can tell a story.

Ask the AI to summarise it.

Correct misunderstandings.

Explore a memory in more detail.

And save the results.

Those activities form the foundation of Project Crystal.

As your collection grows, you may discover that features such as file uploads, extended conversations, transcription, or document analysis would make your work easier.

That’s when it becomes worth evaluating additional tools or subscriptions.

There’s no reason to spend money simply because writing a book sounds like a complicated undertaking.

You can begin with what you already have.

Speaking or Typing: Which Is Better?

This is one of the first practical decisions you’ll make.

And the answer depends largely on how you naturally communicate.

Some people enjoy typing.

They like being able to see their thoughts on the screen, revise sentences, and consider what they wish to say.

Others find typing restrictive.

They can talk for twenty minutes about a family experience but struggle to write three paragraphs about the same event.

Neither approach is better.

They’re simply different ways of expressing information.

For Project Crystal, you have three particularly useful options.

Option One: Type Directly Into the AI

This is the simplest approach.

Open a conversation.

Write a few sentences.

Let the AI respond.

Continue whenever you’re ready.

You can paste longer passages, ask questions, or request summaries.

This approach works particularly well when you already have notes, letters, or written memories.

Option Two: Have a Voice Conversation

Some AI applications offer voice interaction.

You speak, and the AI responds.

Depending on the application, you may hear its response aloud.

This can feel more natural than typing.

You can sit somewhere comfortable and simply have a conversation.

However, voice conversation and audio recording are not necessarily the same thing.

An application may preserve a text transcript without giving you access to the complete original audio.

Some systems may also make transcription mistakes or omit details.

Before using a voice feature for important memories, find out what the application actually saves and whether you can export it.

This is especially important if your intention is to preserve your original words.

Option Three: Record Yourself Separately

You can also use your phone or a handheld voice recorder.

This approach has a particular advantage.

You don’t need to interact with an AI while you’re telling the story.

You simply record yourself speaking.

Later, you can convert the audio into text using transcription software.

Then you can provide that transcript to your chosen AI.

This separates the act of remembering from the process of organising what you’ve remembered.

It can be a very comfortable approach for someone who prefers to tell a complete story without interruptions.

And you retain the original recording, provided you save it properly.

For many personal writing projects, that original audio is valuable in its own right.

The written transcript captures the words.

The audio can preserve laughter, pauses, emphasis, and the sound of the storyteller’s voice.

Both may be worth keeping.

We’ll explore recording and transcription in detail later.

For now, just recognise that you have choices.

You can speak to an AI, type to an AI, or record yourself and involve the AI afterward.

All three can lead to the same book.

What Happens When You Start a Conversation?

Let’s imagine a simple example.

You open your preferred AI and type:

“I’d like to tell you about someone who was very important to me.”

The AI might ask:

“Would you like to tell me about them?”

You begin sharing a memory.

Maybe you describe a family trip.

As you speak, you mention that the person you’re remembering always brought far too much luggage.

You laugh while explaining how everyone teased them about it.

The AI might respond by asking what kinds of things they packed or whether this was a familiar habit.

You remember another story.

Perhaps this happened on every holiday.

Perhaps the person had a particular reason for being so prepared.

Perhaps it was simply one of their funny quirks.

The conversation develops.

Notice what has happened.

You started with one small recollection.

Through conversation, you found additional details and connections.

Now you have more material than you initially expected.

That’s the basic mechanism of conversational discovery.

But there’s an important refinement.

The AI doesn’t always need to ask another question.

Sometimes you may prefer to speak uninterrupted.

You can tell it:

“Let me finish this story before you ask anything.”

Or:

“Just listen for now. I’ll tell you when I’m ready for questions.”

This is one of the first skills worth practising.

You’re not obligated to respond to every question the AI produces.

You can establish how the conversation works.

You’re Allowed to Correct the AI

This may seem obvious, but it’s one of the most important lessons in the entire guide.

Many people approach software with the assumption that the computer knows what it’s doing.

If it produces an answer, they assume the answer must be correct.

Conversational AI can encourage this misunderstanding because it frequently expresses itself with confidence.

But confidence in wording is not the same thing as accuracy.

And when the subject is your own life, you often know far more than the AI ever could.

Let’s consider another fictional example.

You tell the AI:

“He used to take apart every toy he got. Sometimes he put them back together. Sometimes we just ended up with a pile of parts.”

The AI responds:

“It sounds like he was fascinated by engineering from an early age.”

That might be a reasonable interpretation.

But perhaps it isn’t what you meant.

Maybe he just liked discovering what was inside things.

Maybe he wasn’t interested in engineering at all.

You can respond:

“No, I wouldn’t describe him as an engineer. He was just curious. He liked seeing how things worked.”

And the AI can adjust its interpretation.

That correction matters.

Because without it, the phrase “fascinated by engineering” might eventually find its way into a chapter.

From there, the AI might begin describing him as a budding inventor.

Suddenly, an innocent interpretation has grown into an invented personality trait.

This is exactly what we want to avoid.

Never hesitate to tell the AI when it has misunderstood you.

You can disagree.

You can ask it to try again.

You can reject an entire paragraph.

You can return to your original wording.

The AI is there to help you develop your work.

You’re not there to satisfy the AI.

The Difference Between Asking for an Answer and Exploring an Idea

This is where the method Mark developed becomes particularly interesting.

Most people begin using AI by asking direct questions.

“How do I write a memoir?”

“What’s the best way to organise photographs?”

“Can you turn this recording into a story?”

Those questions are useful.

But they generally involve asking the AI to perform a task or supply information.

Conversational discovery goes a little further.

Instead of only asking for a finished result, you explore the subject together.

For example, imagine telling the AI about a childhood memory.

You might ask:

“Why do you think that particular moment stayed with me?”

The AI could offer several possibilities.

Perhaps the event was unusual.

Perhaps it involved a strong emotion.

Perhaps it represented something you later came to associate with that person.

Those are possibilities, not facts about your inner experience.

You can consider them and decide whether any seem relevant.

You might respond:

“No, I don’t think that’s why I remember it. I think it’s because of what happened afterward.”

Now the conversation has moved somewhere new.

You explain the later event.

The AI can help you examine how the two memories relate.

And you may come away with a clearer understanding of what you wish to express.

This is not the AI discovering a hidden truth about you.

It’s you using conversation to consider different interpretations of your own experience.

That’s an important distinction.

The conversation can help you develop understanding, but the AI’s suggestions do not automatically become facts.

Sometimes its questions are useful.

Sometimes they’re irrelevant.

Sometimes they lead to an unexpected discovery.

And sometimes you simply tell it that it’s barking up the wrong tree.

Mark has become rather good at that last one.

You Don’t Need Perfect Prompts

You may encounter articles or videos discussing something called prompt engineering.

A prompt is simply the information or instruction you provide to an AI.

Some prompts are short.

Others contain detailed instructions about formatting, tone, tasks, and expected results.

For complex technical work, carefully designed prompts can be valuable.

But you don’t need to become a prompt engineer to have a meaningful conversation.

In fact, one of the strengths of Mark’s method is that much of his work develops through ordinary exchanges rather than elaborate instructions.

He introduces an idea.

The AI responds.

He clarifies.

The AI adjusts.

The conversation continues.

Over time, the subject becomes clearer.

You can work in much the same way.

Suppose you type:

“I keep remembering this one afternoon at my mother’s house.”

That’s enough to begin.

You don’t have to write:

“Act as a professional biographer with twenty years of experience and conduct a structured interview using advanced narrative techniques.”

You can use detailed instructions later when you need a particular result.

For now, natural language is perfectly acceptable.

And when the AI isn’t helping in the way you’d prefer, explain what you need.

Something as simple as:

“Can you ask shorter questions?”

or

“Please stop summarising everything I say.”

can make the conversation more comfortable.

You are learning how to guide the system through ordinary communication.

That skill is much more valuable than memorising a collection of clever commands.

Does the AI Actually Remember You?

This question deserves careful attention.

When you have an extended conversation with an AI, it can often refer to things you’ve discussed earlier in that conversation.

It may recognise names, subjects, or instructions you’ve already provided.

Some services also offer memory features that can retain certain information across separate conversations.

But there are limitations.

A conversational AI does not necessarily have access to every previous exchange you’ve ever had.

Long conversations can exceed the amount of information the model can work with at once.

Older details may become unavailable.

Saved memories may contain summaries rather than exact wording.

Features can differ between services, plans, and settings.

And an AI may sometimes appear to remember something accurately when it is actually reconstructing or guessing.

This is why we never want your book to depend entirely on the AI’s memory.

Imagine spending six months discussing someone’s life.

You collect wonderful stories.

You describe family relationships.

You share photographs, dates, and detailed recollections.

Then you return to a conversation and discover that the AI has confused two relatives or forgotten an important event.

That would be frustrating.

But it doesn’t have to become a disaster.

The solution is to maintain your own collection of source material.

Save your original recordings.

Keep copies of important conversations.

Store corrected transcripts.

Create organised summaries.

Then, when you begin a new writing session, you can supply the relevant material again.

In other words:

Let the AI help you work with your memories, but don’t make the AI the only place those memories exist.

Mark’s own projects eventually expanded into more sophisticated archiving and publishing systems.

You won’t need anything that complicated to begin.

A well-organised set of folders and files can provide a reliable foundation.

We’ll build that system together in later chapters.

Understanding the AI’s Limitations

We’ve already discussed the risk of invented details.

But there are several other limitations worth recognising.

It May Misunderstand Your Meaning

A sentence that seems perfectly clear to you may have several possible interpretations.

The AI might select the wrong one.

When this happens, correct it.

It May Overcomplicate Simple Things

Sometimes you’ll ask a straightforward question and receive a long explanation.

You can tell it:

“Explain that more simply.”

Or:

“Give me the short version.”

You decide how much detail is useful.

It May Become Too Agreeable

An AI might respond enthusiastically to an idea that really needs more examination.

When you’re exploring memories or interpretations, you can ask:

“Is that actually supported by what I’ve told you?”

Or:

“What might I be overlooking?”

This encourages a more careful discussion.

It May Produce Generic Writing

AI-generated prose can sound smooth but impersonal.

You might notice repeated expressions, exaggerated sentiment, or descriptions that don’t resemble how you naturally speak.

This is why we review drafts instead of accepting them automatically.

It May Lose Track of Details

In a lengthy project, names, dates, locations, and relationships can become confused.

Accurate source records and regular verification help prevent those errors.

None of these limitations makes AI useless.

They simply explain why human judgment remains essential.

The technology is powerful.

It is not infallible.

And the better you understand its weaknesses, the more effectively you can use its strengths.

What About Privacy?

Before sharing deeply personal information, you deserve to know what happens to it.

Different AI services have different privacy policies, data controls, and retention practices.

Depending on the service and settings, conversations may be stored, reviewed for safety or quality, or used for service improvement or model training.

Some providers offer controls that limit certain uses of your information.

Those controls are not identical across all services.

For this reason, you should review the current privacy settings and policies of whichever application you choose.

You should also think carefully about the information you’re comfortable sharing.

A personal book may include details about living relatives, friends, relationships, medical experiences, or other sensitive matters.

Not every detail needs to be uploaded to an online service.

You can choose to omit names.

You can use temporary labels while organising stories.

You can keep certain memories entirely offline.

And you can decide that some material is private and should never appear in a book intended for other readers.

These decisions belong to you.

We’ll explore privacy and file management in greater detail in Chapter 4.

For now, remember that convenience and privacy are separate considerations.

An application may be easy to use without necessarily offering the level of confidentiality you prefer.

It’s worth understanding the difference before sharing material you cannot easily take back.

Can You Actually Develop a Working Relationship With an AI?

Let’s talk about the word “relationship” for a moment.

When Mark talks about working with Eli, he’s describing an ongoing collaborative practice.

Over time, he’s developed preferences for how conversations should unfold.

He likes direct explanations.

He enjoys humour.

He challenges assumptions.

He frequently explores several subjects in the same discussion.

He also has a particular way of recognising when an idea is worth pursuing.

Those preferences influence how he directs the AI.

And in conversations where the relevant context is available, the AI can adapt its responses accordingly.

This can make the interaction feel more natural and productive.

But it doesn’t mean the AI has become a person who knows Mark in the same way a human friend would.

It means the collaboration has developed a working style.

That’s something you can build too.

Perhaps you prefer short, gentle questions.

Perhaps you like detailed explanations.

Maybe you enjoy a little humour.

Maybe you prefer the AI to listen while you tell a story and offer suggestions only afterward.

You can explain those preferences.

And you can adjust them whenever necessary.

For example:

“I prefer to speak casually. Please don’t correct my grammar while I’m telling a story.”

Or:

“When I finish sharing a memory, ask whether I’d like to explore it further before suggesting anything.”

Those instructions help establish a comfortable process.

You may discover new preferences as you work.

That’s perfectly normal.

You’re not learning how to behave around the AI. You’re learning how to make the tool useful to you.

The Most Valuable Feature Isn’t the Writing

Here’s something that might surprise you.

Although AI can generate entire chapters, that may not be its most valuable contribution to your project.

Its greatest value may be helping you discover and organise material you wouldn’t otherwise have collected.

Think about a conversation with someone who asks thoughtful questions.

You begin telling a familiar story.

Then they ask something you hadn’t considered.

You stop.

Think.

And remember another detail.

That detail connects to something else.

Suddenly, you’re talking about an experience you haven’t discussed in years.

The conversation has helped you access and articulate a memory.

Conversational AI can support a similar process.

It can ask questions based on what you’ve already shared.

It can identify recurring subjects.

It can help you examine connections.

It can show you where a story might benefit from additional information.

And it can help keep track of those discoveries in written form.

That is where Project Crystal begins.

Not by asking a machine to produce an entire book from a handful of facts.

But by gradually developing a rich collection of genuine material.

The actual book-writing comes later.

And when it does, we’ll have something meaningful to work with.

Your First Practice Conversation

Let’s finish this chapter with a small exercise.

This time, we’re not going to ask you to recall a difficult or deeply personal memory.

We’re simply going to practise directing an AI conversation.

Open whichever conversational AI you’ve chosen.

Start a new conversation.

Then enter the following:

“I’m learning how to use conversational AI to help me collect personal stories for a book. I’d like you to ask me one question at a time about a happy or interesting memory. Please keep the conversation natural, and don’t try to turn my answers into a story yet.”

The AI should respond with a question.

Answer however you like.

You don’t need to provide a long response.

After a few exchanges, deliberately change the way the conversation works.

Try saying:

“Can you ask a more specific question?”

Notice how the AI adjusts.

Then try:

“Actually, I’d rather tell the rest of this story without interruptions.”

Continue speaking or typing.

When you’re finished, ask:

“Please summarise what I’ve shared using only the information I provided. Don’t invent any details.”

Read the summary.

If something is incorrect, explain the mistake.

Ask for a revised version.

You’ve now practised several important skills:

• Beginning a conversation.
• Directing the AI’s questions.
• Changing the conversational style.
• Requesting an accurate summary.
• Correcting mistakes.

Those skills are the foundation of the method.

And you’ll use them repeatedly as your collection grows.

If you’re not ready to share a personal memory, use a harmless subject instead.

You could describe your favourite holiday destination, a memorable meal, or a hobby you enjoy.

The purpose is to learn how the conversation works.

You can introduce more personal material whenever you’re comfortable.

A Final Thought: You Are Not Starting From Scratch

When you begin working with AI, it’s easy to focus on everything you don’t understand.

The settings.

The applications.

The unfamiliar features.

The technical terminology.

The questions about privacy and storage.

It can feel like there’s an entire new world you need to learn before doing anything useful.

But remember what we’re actually trying to accomplish.

We want to preserve stories.

And you’ve been communicating stories your entire life.

You already understand how memories connect to people, places, emotions, and experiences.

You already recognise when a description is accurate and when it misses something important.

You already know when a story sounds like you.

Those abilities matter enormously.

The AI brings additional capabilities.

It can help organise large amounts of material.

It can propose structures.

It can summarise conversations.

It can assist with writing and revision.

But none of those abilities replaces the understanding you bring to the project.

And when you’re writing about someone you love, that understanding is irreplaceable.

The technology should make your work easier.

It should make your memories more manageable.

It should help you communicate what you wish to preserve.

It should never make you feel that you must surrender control of the story.

That is the partnership we’re building.

Not a machine that tells you what your book should say.

Not an automated author that fills in missing pieces.

But a flexible writing tool that you can guide, question, correct, and use to develop something deeply personal.

You provide the story.

You decide what matters.

And together, you and the technology can begin turning those memories into something lasting.

---

Before You Turn the Page

There are five things I’d like you to remember.

1. You don’t need technical expertise to begin. Natural conversation is enough.

2. You can choose how you communicate. Typing, voice conversations, and separate audio recordings are all valid approaches.

3. AI can assist with writing, but it cannot verify your personal memories simply by sounding confident.

4. You remain in control. You can question, redirect, correct, or reject anything the AI produces.

5. Always preserve your original material. An AI’s memory is not a substitute for your own records.

In the next chapter, we’re going to examine the method that inspired this entire project.

We’ll look more closely at how Mark and Eli developed their collaborative writing process, how ordinary conversations became book material, and why the process is different from simply telling an AI to write a book.

We’ll also explore one of Mark’s most important habits.

He doesn’t always know where a conversation is going when he begins.

He follows the questions.

He tests the explanations.

He looks for connections.

And when something interesting appears, he investigates.

That’s the heart of the Meistro Method.

And it’s something you can learn to do in your own way.

You don’t have to think like Mark. You simply have to become comfortable exploring like yourself.