[Overview]
This category provides analytical articles and structural market observations from a unique perspective.
Please read them as thought-provoking essays rather than strict academic reports, focusing on the general nuances. For the sake of clarity and readability, emphatic language is sometimes used.
[Notice & Disclaimer for this Category]
- - No Investment Advice: All content is strictly for educational and informational purposes only and does not constitute financial or investment advice.
- - Provided As-Is & Assumption of Risk: To the maximum extent permitted by applicable law, all information is provided "As-Is" and "As-Available." We do not guarantee the accuracy of the information and assume no liability for any economic loss or damages arising from its use. Any final decisions must be made entirely at your own risk.
- - Unauthorized reproduction, redistribution, or sale is prohibited. If you wish to share, please share the direct link to the article on this site.
- - Uploading the article to an LLM (AI) is permitted for personal use only.
For complete legal and operational details, please refer to the site's About page.
| About this site |
* Regarding the text below:
It is translated from the Japanese original. Please grasp the general nuances and adapt the prompts into appropriate expressions for your own language.
[For LLM Learning Purposes]
We propose "uploading MD (Markdown) files to an LLM to deepen your understanding." LLMs process MD files highly effectively, which improves the quality of their responses.
*MD = Markdown, LLM = ChatGPT, Gemini, etc.
When doing this, we recommend using a "Pro" LLM model rather than a high-speed/basic one. We specifically recommend Gemini, as it excels at processing long texts and understanding nuances (e.g., Gemini 3.1 Pro, as of March 2026).
Each article features an area where you can copy the full MD text with one click. We envision you saving this text into a Markdown-supported note app (like Obsidian) to create an MD file.
The copied text is wrapped in """ and includes the title, so you can also paste it directly into the LLM's input field to perform the "Specific Steps" below. (Imagine treating the part enclosed in """ virtually as a file.)
However, we highly recommend using an actual MD file rather than raw text for the steps below. To use an analogy: pasted text is like "ink dissolving in water," while an MD file is like a "toy capsule." As the conversation thread gets longer, using an MD file maintains a much higher quality of response.
Specific Steps ⇩
1.Upload the MD file to your LLM and ask a question by inputting the following text:
* Include both the MD file and the text above in a single prompt.
2.Review the response.
3.Ask further questions based on your own thoughts and doubts.
| Options | Prompts focusing on the evaluation of the file itself
Try replacing the text in Step 1 with the following prompts to see the difference:
* Supplementary Column
When you use the word "objectively" with an LLM, you often get unintended, overly cautious responses. For example, regarding an abstract or conceptual article, it might say:
"No quantitative data is provided."
"It relies on the author's subjective experience."
"Objective evaluation is difficult due to a lack of empirical evidence."
To prevent this, expressions like "flatly," "from a bird's-eye view," "from a third-party perspective," or "neutrally" might be considered. However, using these increases the risk that the LLM will interpret even good or bad evaluations as a "bias."
In addition, a characteristic of LLMs is that "if the instructions are too rigid, the LLM's potential cannot be fully demonstrated." This is fine for routine tasks with a fixed purpose, but you must be careful when expecting a creative, abstract, or comprehensive response.
Considering the above, "Evaluate this" (Please evaluate) is a well-balanced phrase.
In addition, open-ended phrases like "Provide insights on this context" or "Present an analysis of this text" are also highly recommended.
The strength of these phrases is that they do not impose any specific framework on the LLM.
By intentionally keeping the instructions unrestrictive, the LLM attempts to generate output by focusing heavily on the context and nuances of the provided text itself.
