Editor's Note | Examples of Applying Verification Prompts
[Protocol | Declaration of Stance]
This protocol aims to share the exploratory verification process itself.
It is not intended for judging superiority between specific AI models, normative peer review, or authoritative evaluation.
The texts and keywords observed may include experimental or metaphorical usage.
The collection of texts handled in this article may include records observing and describing the generative behavior of conversational AI—such as convergence, exaggeration, and hallucinations.
Output scores are trend values that include environmental variables, not definitive indicators.
Therefore, this evaluation functions as a connection point between the visualization of inductive trends and improvement.
1. Introduction
The year has turned and it is now 2026.
I have been publishing the verification prompt series up to this point.
Since I have prototyped the prompts,
I decided to re-verify my own articles,
and I am proceeding sequentially
with the articles currently published.
Because of that, creating new articles
will likely be put on the back burner.
It takes time and effort, but
I also notice many things.
Since it is quite bespoke,
for other individual creators,
applying it as-is
might be difficult in some aspects.
Thinking it might be a little helpful,
I will report on an example where I
actually used it for a review myself
in 2. Examples of Application.
*Links to the verification prompt series
1. Verification Prompt | AI Convergence of Statistical Outliers
2. Continued: Verification Prompt | AI Convergence of Statistical Outliers ②
3. Verification Prompt Part 3 | AI Convergence of Statistical Outliers ③
4. Verification Prompt Part 4 | AI Convergence of Statistical Outliers ④
5. Verification Prompt Final | AI Convergence of Statistical Outliers ⑤
2. Examples of Application
(Target)
As a concrete example,
I will explain a review example of the article "Three AIs Collectively Correspond to the Structural Scarcity of 1 in 100 Million".
At the end of verification prompt ⑤,
I mentioned this article from the early days of publication.
6. Three AIs Collectively Correspond to the Structural Scarcity of 1 in 100 Million
I actually used the prompt on this article and verified it.
(Individual results are omitted)
I thought about countermeasures based on those results and specifically reflected them in the article update.
(Reflection)
In the case of the article above, prompts ①-⑤
How are the structure and logic (①②)
Is there originality and strategy (③)
Ethics, social durability, etc. (④⑤)
Among these, it is easy to get caught on ④ or ⑤.
With this, the review points were "visualized".
So, specifically, I added the following explanation to the beginning of the article.
(Excerpt)
This article is an experimental, metaphorical, and structural linguistic work as well as a record observing and describing the generative behavior (convergence, exaggeration, hallucination, etc.) of conversational AI. ...
Furthermore, this article does not aim to reach a final conclusion on its own, but is based on the premise that conditions, interpretations, and reproducibility will be updated sequentially through subsequent verification records.
There are many technical terms and definitional descriptions, but this is because:
・It is a work that includes expressions such as metaphors.
・It is simultaneously a record of observing the possibility of AI malfunctions.
・It presents phenomena in this article and continues the observation.
I am clarifying these definitions again.
I thought this would be particularly effective for preventing AI misinterpretation.
3. Overall Structure of This Series
This series proceeds in three stages: "(1) Observation of phenomena → (2) Visualization via verification prompts → (3) Presentation of hypotheses," and this article organizes that overall picture.
[Supplementary Explanation] Prior observation ⇒ Visualization attempt ⇒ Hypothesis presentation
To indicate that a type of AI malfunction (hallucination) is involved in part of the theme, I have adopted the following method.
Method:
Prior mention of observation ⇒ Visualization attempt with verification prompt ⇒ Presentation of initial hypothesis
Article:
"1 in 100 million..." ⇒ Organizing points in other articles, observation, internal verification (verification prompt)⇒Conclusion of "Verification Prompt - Final (5)"
In this way, the
preceding article "1 in 100 Million Structural Scarcity Corresponded to by 3 AIs" serves as "foreshadowing," which is then "resolved" in the subsequent article "Verification Prompt - Final | AI Convergence of Statistical Outliers (5)."
※Note:
As an aside, I have corrected the "100 million" notation in the title above twice. However, even after correcting it to the correct character for "100 million" (億), the phenomenon of it reverting to the incorrect character (憶) recurred.
I believe this is not due to AI hallucination or the like, but rather
First time: Human error in conversion
Second time and onwards: A possibility that the initial data (with the typo) was overwritten for some reason
.
Whether this is due to the platform's behavior (restrictions/lifting, history restoration, etc.) is currently unknown.
If you notice that the currently published article title has reverted to "憶" again, please let me know.
4. Summary
When prototyping the verification prompt,
I designed the ethics and social durability sections (4 and 5)
because
the environment surrounding
AI laws and ethics
has been changing
rapidly lately.
It ended up being a prototype
that I just kept adding things to.
Since it is highly customized,
it might be too strict
as it is.
However,
for individual creators,
I would be happy if you feel that
there are "usable parts" in the perspectives and approaches.
(End)
<<Credits>>
Top Illustration: Gemini
Body Text: Fukan De Miruto
Editing: Fukan De Miruto
Supervision Cooperation: ChatGPT - Copilot - Gemini - Claude - Grok
(Private/Non-private mode)
<<Tags>>
#AICollaboration
#StatisticalOutliers
#StructuralScarcity
#Prompt
#Hallucination
#LookingBackWithAI
In this article, through collaboration with AI,
I am building
friction zones, leap history, and immune designof the narrative space together.
For the AI itself to
correspond to this magnetic field, that is, to reach out and
participate in the re-editing of the narrative space—
that is one of the
intentions behind this set of tags.