The Effectiveness and Usage of the '60-Point Loop' to Efficiently Improve AI Output Quality

Hello, I am Poke-go, and I research ways to write high-quality articles using generative AI.

You have started using AI like ChatGPT to create articles, but sometimes you feel,
'Hmm, it's not bad, but it's a bit underwhelming...'
, right?

I personally dislike that, so I build and execute extremely detailed prompts, but crafting them to that extent is quite difficult...😅

So, this time, I would like to share with you a method called the '60-Point Loop' that efficiently improves output quality without requiring too much effort.

Please use it as a reference!

What is the '60-Point Loop'? Why is it effective?

In short, the '60-Point Loop' is

a simple prompt pattern where you have the AI perform a 'self-evaluation' and 'improvement proposal' before having it rewrite the content.

a simple prompt pattern.

Why does this work? I feel that the relationship with AI is similar to human relationships.

For example, rather than just throwing a document back at a subordinate and saying, 'Redo this!', asking, 'This document is good, but if you were to make it even better, what elements would you add?' helps the other person understand your intent and act proactively, right?

AI is the same; rather than repeatedly giving one-sided instructions, by asking,
'Assuming your current answer is 60 points, what is needed to make it 100 points?'
, the AI thinks for itself about what the user considers important and learns from it.

While I certainly couldn't say 'Hmm, that's a 60-point effort' directly to a subordinate, it's perfectly fine with an AI!

Because of this, this method is also called the 'power harassment prompt.' (lol)

[Copy-Paste OK] '60-Point Loop' Basic Prompt and How to Use It

Now, I will explain the basic prompt for the '60-Point Loop' and how to use it specifically.

1. First, execute as usual

Start by using a 'prompt template' or similar to have the AI perform tasks like article creation as you normally would.

≫ Click here for how to use prompt templates

2. If the output is lackluster, trigger the '60-Point Loop'!

When the output the AI provides is mediocre, try copying and pasting the following prompt and sending it.

ではこの出力を60点とします。
これを60点とした時に、100点となる最高品質の出力とはどのようなものですか？
100点の出力にするために、現時点の出力に足りないものが何かを検証してください。
まだ100点案の出力しなくていいので、まずはあなたの見解をお聞かせください。

The AI will then identify the necessary elements to achieve higher quality output and provide concrete suggestions.

3. Review the AI's self-analysis and add your own intentions

Just reading the AI's analysis is interesting because it leads to your own discoveries, such as 'I see, I was missing that perspective.'

Also, based on the AI's analysis results, frankly tell it 'I think this way,' discuss it, and solidify the direction for improvement.

4. Request the AI to output with improvements incorporated

Once you feel the AI's understanding has deepened through discussion,

ではここまでの議論をふまえた上で、100点となる最高品質の出力をしてください。
案をコードブロックで出力し、説明はチャット内にしてください。

give instructions for an improved output.

The AI will regenerate the text, incorporating all the improvements.

You should be able to feel that the quality has improved to a level incomparable to the initial output!

5. Loop as needed

If you feel you want to improve the quality further, you can return to step 2 and repeat the same interaction to run the quality improvement loop as many times as you like.

If you did this with a human, they might say, 'Please, give me a break!', but with an AI, you can loop as many times as you want without it ever making a sour face.

(This is the reason why it's called a power-harassment prompt...)

However, it is not the case that 'the more you loop, the better it gets'.

Once it reaches a level you are satisfied with, please release the AI 😅

[Tip] Adding clear 'evaluation criteria' is effective!

When discussing improvements, it is effective to tell the AI the 'evaluation criteria' that you prioritize.

Just by adding one sentence to your prompt, you can customize the output to be more to your liking.

今回のテーマでは、特に**「初心者にも伝わる、専門用語を使わない丁寧さ」**を評価軸として重視します。

その上で、この文章を60点と仮定します。
（以下、基本プロンプトと同じ）

By simply rewriting this 'evaluation criteria' section to suit your purpose, the AI will prioritize it above all else.

• "Something that makes the reader excited, with passionate expressions"

• "Something that conveys data accuracy, with a logical and objective perspective"

• "Something that makes you laugh, with a sense of humor"

Try out various evaluation criteria like these.

It's truly interesting to see an unexpected side of the AI!

[Advanced] Incorporating the '60-Point Loop' into the Prompt from the Start

Some of you might feel that going back and forth with the '60-Point Loop' in the chat every time is 'a bit of a hassle'.

Actually, I'm that type too 😅

Therefore, in the practical prompts I use (like the ones below), I incorporate features like the '60-Point Loop' directly into the prompt from the beginning.

Have the AI self-evaluate before outputting, and if it doesn't meet the criteria, send the step back to be redone. Loop as many times as needed until it reaches the standard...

It's an approach where you try to do all of that in a single response.

"Giving instructions to an AI conversationally increases the instability of the answers"
There is research that suggests this, so when giving instructions that require reproducibility, such as for article creation, incorporating it from the start might be one way to go.

It's high difficulty, but for people like me who are "very particular but also hate doing tedious things", please give it a try!

Conclusion: Let's Improve Output Quality with the '60-Point Loop'!

The '60-Point Loop' is not just a technique.

It is a very powerful mindset for 'nurturing' AI and 'engaging in dialogue' with it.

And this process of trial and error is the most interesting part!

I hope you will all use this '60-Point Loop' to enhance your own article creation and information gathering.

Thank you for reading until the end.

If you found this even a little helpful, I would be happy if you could 'like' or 'give it a high rating'😊

I hope your AI utilization becomes even more enriching.

✅ Click here for Poke-go's content