What Changes When You Give AI a 'Thinking Pattern'

Last time, I talked about how to stop wavering in your decision-making by "placing your procedures outside your head."

This time, I will continue that discussion. I will write about what happens when you not only place your procedures outside, but also share them with AI.

The true nature of "not being able to use it well"

If you use AI on a daily basis, you have probably had this experience.

Even though people are using the same tools and the same features, one person says it is "incredibly useful," while another feels it is "not very useful."

Regarding this difference, what is often mentioned is "how to write prompts." The idea is that if you improve the precision of your instructions, the output will also improve.

That is correct, but there is a deeper, real problem.

Whether your thinking is organized before you write the prompt.

A prompt is ultimately just a language for conveying your thoughts to AI. If the content you are conveying is vague, no matter how carefully you write it, it will be a vague request. In many cases where people feel they "cannot use it well," the problem is not the prompt, but the degree to which they have organized their own thoughts before writing the prompt.

What does it mean to pass a pattern?

When you say "passing a thinking pattern to AI," it sounds difficult. But what you are doing is simple.

You convey the "procedure" you use when you think to the AI in advance.

For example, suppose you are writing a proposal. The procedure I always use to think is as follows: First, organize the other party's situation. Next, identify the problem the other party truly wants to solve. Then, translate what I can offer into the other party's language. Finally, show only one action the other party should take next.

If you pass this procedure to the AI in advance, the AI will operate within that procedure. What comes out is fundamentally different from when you just give the instruction "write a proposal."

Passing a pattern means giving the AI a "map of how to think." Without a map, even the best navigator will get lost. If you have a map, you can continue to move in the same direction without having to give instructions for every single judgment along the way.

What changes and what does not change

Let's look specifically at what changes when you pass a pattern.

First, the stability of the output increases.

Without a pattern, AI output depends heavily on the "quality of the input for that day." This is the same as what I mentioned last time: if the input wavers, the output wavers. However, when there is a pattern, even if the input wavers slightly, the skeleton of the output remains stable. Because the filter of the pattern is applied, the wavering is absorbed.

Next, the cost of revisions decreases.

When you feel like something is 'a bit off' without a pattern, it is difficult to articulate exactly what is wrong. You have a vague sense of discomfort, but you cannot pinpoint where or how it differs. In this state, correction instructions often miss the mark, and time just slips away. With a pattern, you can point out, 'The angle of this part in this step is wrong.' Because the granularity of the correction becomes clear, the number of exchanges decreases.

On the other hand, there are things that do not change.

The final decision is made by you.

By providing a pattern, the AI structures, proposes, and organizes according to the procedure. However, it is you who decides, 'I will go with this.' The AI plays the role of preparing the ground for the decision, but it does not make the decision itself.

This is not a limitation, but a design premise. As long as the subject of the decision remains with you, no matter how much you utilize AI, your own thinking will not be lost. With this premise, it is not dependence on AI, but collaboration with AI that is established.

On the word 'reproducibility'

The greatest advantage of providing a pattern actually lies in 'reproducibility'.

Without a pattern, it is difficult to reproduce a task that went well the next time. Because 'what I was thinking and how I gave instructions to get that result' has not been verbalized, you end up relying solely on memory.

With a pattern, the procedure that worked remains as it is. If you bring the same pattern to the same type of task, you will get an output close to the same quality. You can use the same procedure whether you are working alone or in a team.

The word 'reproducibility' has a somewhat manufacturing-like ring to it. But in intellectual work as well, the same thing should fundamentally hold true. Verbalizing the methods that worked and making them reusable. I have often thought recently that this is what 'improving the quality of work' means in its original sense.

The concept of a 'Thinking OS'

From here on, I will talk about what I have actually been working on.

Placing procedures outside, handing them to AI, and ensuring reproducibility. When I tried to create this mechanism, I naturally arrived at the concept of an 'OS'.

In the computer world, an OS is the 'foundation upon which all applications run.' No matter how high-performance individual apps are, if the OS is unstable, the whole thing will not work well. Conversely, if the OS is solid, the things running on top of it will be stable.

I thought that thinking might be the same. Writing emails, planning, analyzing, explaining to others. These are each different 'apps,' but there is a 'decision procedure' common to all of them. If that common part can be externalized as an OS, individual tasks will become stable.

I call this way of thinking a 'Thinking OS.' And now, I am providing a mechanism that has been refined into a form that can be used with AI as an actual service.

However, in this article, I wanted to convey 'why the concept of a Thinking OS itself is effective' more than talking about that service. This is because, regardless of whether you use the mechanism or not, the idea of 'placing procedures outside' is something that everyone who uses AI can use right now.

Next time, I will write about how this 'Thinking OS' was born and the awareness of the issues behind it. In particular, it will be a slightly personal story about the feeling of 'being replaced by AI' and how I came to terms with it.

In this series, I will write about the 'cognitive wall' that everyone hits when collaborating with AI, and how to overcome it.