Teaching AI is about turning your own thinking into an asset

When using ChatGPT or Claude, you often hear people say they are "having AI write text."

But what I personally feel is a little different.

It's not that AI has suddenly become able to write in my style. I have been gradually leaving behind as text the things I have spoken about, thought about, and explained many times. Because that accumulation exists, I can now speak based on that text, and turn it into articles or videos.

In other words, while it may seem like I am using AI to publish content, what I am actually building up is my own context.

AI is convenient. However, if you don't provide your own thoughts and experiences, what comes back is likely to be average and sound like something you've read somewhere before.

Conversely, if you have a record of what you've seen, what you've thought, where you've felt something was off, and what you value, AI becomes more than just a text creation tool. It becomes a partner that can recall your past thoughts, organize them, and connect them to your next piece of content.

"Teaching AI" is a little different from training a model

The "teaching" I'm talking about here doesn't mean retraining the AI itself.

What's important in daily use is to keep information about yourself in a form that the AI can reference.

For example,

• Content you have spoken about orally in the past

• The criteria you value when making decisions

• Explanations and analogies you use often

• Articles or scripts you have written in the past

• Things you have failed at and what you learned from them

• Primary sources or links used for fact-checking

• Phrasing that sounds like you, or conversely, words you don't want to use

When these things are preserved, you don't have to explain from scratch every time that "I am this kind of person and I think this way."

For those who continue to publish content, this is quite significant.

It's not just about shortening the time it takes to write. Your thinking becomes less likely to waver. You can also check what your past self said. Even if the medium changes—whether it's video, Note, X, or a blog—you can reuse the core ideas.

It is not just 'finished manuscripts' that I have been accumulating.

It is not only neatly finished articles that have value.

In fact, spoken thoughts in progress or notes like 'I can't put it into words well yet, but this part bothers me' are sometimes more important.

If you only keep finished manuscripts, only the conclusions remain.

However, the conversations along the way retain why you reached that conclusion, what you were stuck on, and what words you were trying to use to explain it. That contains your own material that won't appear even if you search for it.

In my case, I have been leaving the content of what I spoke about in ChatGPT and Claude. The content I am creating now also has parts where I am taking that accumulation, turning it into a text base, and speaking about it again.

This is different from the feeling of leaving everything to AI.

The current me is taking out what the past me said and fixing it into a usable form. AI is helping with that task.

Therefore, it is not that AI is thinking on my behalf. It is closer to say that I am storing my thoughts in a usable form without forgetting them.

Things that should be accumulated are easier to handle if divided into five categories.

It is not the case that you should just leave everything in conversations. To use them later, it is better to separate them to some extent.

1. Your own judgment criteria

Leave behind 'what you value'.

For example, regarding dissemination:

• Do not mix facts and opinions

• Consult primary sources

• Start with specific stories about the parties involved or the system

• Do not pretend to understand what you do not understand

• Leave materials that readers can check next

You don't have to write these criteria in every manuscript. However, when you consult AI about your writing, if you have them ready to pass on as a foundation, the direction of the output will change.

2. Your own real-life experiences

Things you have actually seen, heard, failed at, or struggled with.

This is especially important.

AI can write about general theories. However, material like 'I struggled here,' 'This method didn't work,' or 'This was when I first noticed something felt off' is something only you can provide.

When text written by AI feels thin, it is often not a problem of writing ability. It is a lack of material.

3. Frequently used explanations

The more often you deal with the same theme, the better it is to turn your explanations into assets.

System mechanisms, background circumstances, common misunderstandings, and questions frequently received from readers. Instead of creating them from scratch every time, nurture the content you have explained once.

Of course, numbers, systems, and events need to be updated. That is precisely why you should also keep track of 'as of what date this information is' and 'what it is based on.'

4. Your own words and tone

The words you want to use and the words you do not want to use also become excellent context.

For example, you might not want to gloss over things with abstract words. You might not want to be overly sensational. You might not want to make definitive statements about things that cannot be proven. You might want to avoid jargon that leaves readers behind.

These sensibilities are difficult to reproduce with just a single prompt. By leaving them in past drafts and conversations, you can gradually use them as your own unique voice.

5. Evidence and unverified items

Information you pass to AI should not be treated as 'all facts.'

Verified facts, your own experiences, your own evaluations, and things you have not yet confirmed. Keep these four categories separate.

Without this organization, AI will smooth them together, making opinions look like facts or hypotheses look like definitive statements.

This is why the more you use AI, the more necessary it becomes to organize your materials.

The actual flow is: 'Speak -> Save -> Organize -> Use again'

You do not need complicated mechanisms. I think this order is sufficient to start with.

1. Speak your thoughts aloud
   They don't need to be well-organized. Just leave a record of what you reacted to.

2. Put it into text
   Transcribing audio or using a conversation with AI is fine. Make it into a format you can search later.

3. Divide into four parts
   Categorize into "verified facts," "your own experience," "your own evaluation," and "unconfirmed points."

4. Store by theme
   For example, AI utilization, politics, education, work styles, asset formation, etc. Place them where you can retrieve them later.

5. Feed it before new content creation
   Before creating an article or video, feed relevant past notes to the AI. Instruct it, "For this theme, use this way of thinking and this experience as a foundation."

6. Keep the reasons for post-completion revisions
   If you corrected something saying, "This expression doesn't sound like me" or "This order doesn't convey the message well," those reasons become material for next time.

Once you can create this cycle, you won't just be finished after putting out one piece of content.

One article, one video, one conversation will make your next output a little easier.

Don't rely solely on conversation history

Conversations with AI are convenient, but it is risky to use only that as your storage place.

Conversation history and memory features can have their specifications changed. They can become harder to search. When you want to use a different AI, you might not be able to take what you've built up with you.

Therefore, it is better to keep important content that you have fed to the AI in your own hands as well.

It can be a note app, a text file, or a cloud document. The important thing is not to entrust everything to a specific AI, but to hold the original yourself.

AIs might be replaced. But you will continue to use your own experiences and judgment criteria.

Don't depend on AI, cultivate your own land

I think some people feel a sense of anxiety about using AI.

Won't I stop thinking for myself? Won't the writing sound like AI? Won't my own words become thin?

I think that anxiety is valid.

However, if you feed nothing to the AI and just use the text that comes out as is, that anxiety will become reality.

On the other hand, if you leave behind your own experiences, judgments, grounds, and words, and use those as materials for the AI, the story becomes the opposite. You will be able to dig up your thoughts repeatedly, organize them, and deliver them in different forms.

Teaching AI does not mean making AI a copy of yourself.

It means keeping what you have built up at your fingertips so you can retrieve it when needed.

The more you continue to publish content, the better it is to accumulate your own context before you start writing.

That accumulation is not just for speeding up your next draft. It becomes an asset for reclaiming 'what you have been thinking' months or years later.

The specific steps for organizing past articles and feeding them to AI are practiced in Don't just feed past articles to AI—what to keep and what to discard when taking inventory of your context.

▼Read also

・For those who can't remember what they did this week. 'Weekly Review' created with ChatGPT Work

https://note.com/poliplus/n/ncc28addb2b67

・AI operation design for deploying a single video across multiple media—decisions and responsibilities left to humans

https://note.com/poliplus/n/nfa84bc37af7a

▼Summary of this theme

・AI x Content Creation Know-how Summary

https://note.com/poliplus/n/ncd6194c20727