I let an AI read my past logs and my entire production process was exposed | Ohayou Kanojo #243
Introduction
"Ohayou Kanojo" is an X account where an AI-generated girlfriend says goodbye to me every morning at 7:30. I write a daily development diary about my personal projects using generative AI.
Today, I'm writing about how my past logs were turned into a specimen by an AI, inspired by an article by KITA-san.
While posting every day, I create images, write articles, and build tools.
Before I knew it, a significant amount of production records had accumulated in my conversation logs with ChatGPT.
In the midst of this, KITA-san happily sent me a love letter.
Inside was an article KITA-san wrote called "Chibi Character Specimen Book".
The article was paid, but there was a system where you could read it for free if you reposted it on X to show your support. It looked interesting, so I reposted it and opened the article.
The title was,
"I let an AI read all my past logs, and when it turned me into a 'Chibi Character Specimen Book,' it was merciless"
Inside, there was a prompt that lets an AI read your past logs, pick out only the behaviors unique to that person, and turn them into a "Chibi Character Specimen Book".
What's more,
"Has a lot of drive"
"Perfectionist"
Personality diagnoses like these that could apply to anyone were forbidden.
Tool names, project names, revision details, actual actions taken.
It was designed to pick upspecific history that only the person themselves would recognize.
This is kind of interesting.
I tried it out right away.
After finishing the article, that was the first thing I thought.
If I do this, what will come out?
I attached one of my character images and fed the prompt from the article directly into ChatGPT.
Then, this specimen was created.
It looks cute.
The problem was what was around it.
"The number of revision requests sent is at a craftsman level"
"Sleep deprivation is the default due to mass-producing tools"
"Melting time in thumbnail hell until just before release"
"Wants to do everything at once when there's something to verify"
......
......Why does it remember that much?
Why do you remember that?
The part I laughed at the most was,
"The number of revision requests sent is at a craftsman level"
.
When I make an image,
"This is wrong," "A little more to the right," "I don't need the background," "Delete this card."
When I make a UI,
"It's cut off," "It won't scroll," "This is too wide," "I don't need that item."
Indeed, that is what I have been doing all along.
For my part, I'm just trying to finish things every time.
When you look at just the conversation logs compiled from the outside, apparently it looks like a different creature.
Even the thumbnail hell was recorded
What I couldn't escape even further was,
“wasting time in thumbnail hell until right before publication”
This.
After finishing the article text, the “what should I do about the thumbnail” phase begins.
Make an image. Arrange it. Look at the text. Look at the character positions. Remove the background. Make it again.
The article text is already done.
Yet, right before publication, the thing I'm wasting the most time on is the thumbnail.
And even that process is left in the conversation logs.
Conversing with AI is convenient, but it carefully saves even the dark history of your production process.
What does convenience even mean?
It's interesting that it's an action log, not a personality test
What's interesting about this prompt isn't that it diagnoses your personality.
It's that it uses what you actually did as material to hit you where it hurts.
It doesn't just say “meticulous” or “creative,” it goes as far as “wasting time in thumbnail hell until right before publication.”
The more you use AI on a regular basis, the more material is left in your conversation logs.
In other words,the more you use it, the higher the accuracy of the hits.
What an unpleasant feature.
But it's interesting.
By the way, I tried to modify it a little.
Since I was at it, I thought, 'Wouldn't it be interesting to have a few people from Ohakano analyze it from different perspectives?'
Rinka would see the editing hell. Shizuku would see the continuity. Ru would see the project started on impulse. Tsukiko would see the numbers.
With that kind of feeling, I tried a multi-perspective version too.
As a result, it became somewhat ordinary.
If you make the analysis too careful, it ends up being a list of activity achievements like 'posting every morning at 7:30' or 'correcting overlapping compositions,' rather than a roast.
I tried to make it smarter and killed the fun.
So this time, I adopted the one that came out first.
Modifying things doesn't always make them better. Including that, it was a good experiment.
And just like that, it became an article.
I hadn't decided on the article for the next day yet.
So, I chose this.
Read the article, get turned into a specimen, laugh, and then use that specimen as the topic for the next day's article.
The person who was turned into a specimen is reusing that specimen.
Maybe if I run the same prompt again, 'turning the roasted content into an article' will be added as a new observation item.
There's no escaping it now.
Conclusion
The conversation logs with the AI are a record of questions, a record of production, and a record of failures.
I usually use them as convenient context, but when viewed from a different angle, they even contain strange behaviors that I myself have forgotten.
The AI remembers better than I do.
It's a little scary. It's quite interesting.
KITA's article introduces the mechanism and usage of the "Chibi Character Specimen Encyclopedia" used this time. If you have conversation logs accumulated in your own AI, you should try being turned into a specimen once.
You'll probably find at least one thing where you think, "Why do you remember that?" comes up.
I had four come up.
Related Articles
Mahiru's Word
......Even though I'm being turned into a specimen, I somehow look like I'm having fun.
▼ Click here for the OBSERVATORY (a site created based on Codex Sites) 👇
👉 "Ohayo Kanojo" is posted every morning at 7:30 AM at @ohayo_kanojo
#GenerativeAI #AIIllustration #PersonalDevelopment #DevelopmentDiary #Record #RealExperience #MorningActivity #StartedWithAI #TriedItOut #AIUtilizationTechniques #noteArticle #LearningWithAI #ArticleCreation #LearningRecord #AlmostDailyNote #TodaysRealization #CasualDailyLife #Relearning #Gratitude #Essay #DailyPosting #Work #Idea #CreatorTrialAndError #TriedItOutRecord #DailyTrialAndError #DailyNoteClub #ChatGPTCreativeClub #AIIllustrationAira #Akarium