I turned 185 articles into 5 note analysis reports | Ohayou Kanojo #185

Introduction

"Ohayou Kanojo" is an X account where an AI-illustrated girlfriend says goodbye to you every morning at 7:30. I write a daily development diary about my personal development using generative AI.

Reading KOTAYA-san's analysis article really struck a chord with me.

Analysis can be delivered not just through text, but through images as well.

Until now, when I looked back on my note activities, I organized them using numbers and text. But I never had the idea of "making it visible at a glance."

However, I didn't want to make the same thing. I wanted to experiment with redesigning it using my own data, my own characters, and my own observation logs.

So this time, I created 5 infographics based on 185 articles. This is the production log for that.

Composition of the 5 images

I assigned a "day of the week" character from "Ohayou Kanojo" to each of the 5 images.

The characters are not the main focus, but rather exist to guide the analysis.

Image 1 "Account Summary" Person in charge: Hiyori
A page that shows who I am and how long I've kept it up.

Image 2 "Monthly Growth" Person in charge: Tsukiko
A page that shows how it grew and where the turning points were.

Image 3 "Brain Map" Person in charge: Shizuku
A page that shows what I am connecting and how.

Image 4 "Category Analysis" Person in charge: Rinka
A page that shows what I am writing about and what generates reactions.

Image 5 "Reader Map" Person in charge: You
A page that shows who I am building relationships with.

Opening with Hiyori and closing with You. Even though it's an analysis, it ended up being a composition where human warmth returns at the end.

Image 1: Account Summary (Person in charge: Hiyori)

Hiyori is in charge of "who I am and how long I've kept it up."

A total of 185 articles, a posting frequency of 1.05 posts/day, and 0 missed days. A cumulative total of 7,271 likes and 2,270 comments. Data from about half a year since the start of activities is contained on one page.

The most prominent figure among the numbers was "0 days missed."

I thought the fact that it "never stopped once" was stronger information than the continuation itself. Assigning Hiyori was the right choice, as it resulted in a layout that gently wraps around the content.

Image 2: Monthly Growth (Assigned to: Tsukiko)

Tsukiko is in charge of "how it grew and where the turning points were."

When lining up the number of articles, average likes, and average comments per month for 6 months, a major shift appeared in March.

Since Tsukiko is in charge of organizing data and timelines, she was a good fit for the growth page. One mistake was that the label in the image ended up being "A word from Hashamo." It should have been "A word from Tsukiko." I will fix it in post-processing.

Image 3: Brain Map (Assigned to: Shizuku)

Shizuku is in charge of "what is connected to what and how."

Six clusters—emotions/observations, character design, tool development, continuous operation, relationships with readers, and experiments/expansion—radiate outward from the central "Hashamo."

This image was the most prone to breaking. When instructed to create a "brain map," AI immediately tries to make a collection of words (word cloud). When I explicitly stated, "Place them as small rounded cards and connect them to the central node with lines," it properly became a structural diagram.

Image 4: Category Analysis (Assigned to: Rinka)

Rinka is in charge of "what is being written and what generates reactions."

I divided the 185 articles into 5 categories: tool development, character operation, design philosophy, creative experiments, and mini-logs.

The thing I was most careful about here was not putting numbers in the donut chart.

The roles of the categories were organized, but the ratio of the number of articles had not yet been finalized. If you leave it to image generation, it will calmly create numbers like "28%" or "31.4%." If you put baseless numbers in an image, it becomes a correction nightmare later.

In the center of the chart, I only placed role labels: "Most frequent: Tool development / Foundation: Character operation."

Image 5: Reader Map (Assigned to: Yo)

What I, Yo, am in charge of is 'who is developing a relationship'.

967 comment bodies, 164 unique commenters. It would have been easy to create a 'Reader Ranking Top 10' from this. But I didn't do that.

When you rank readers, the quality of the relationship disappears.

People who comment a lot, people who quietly support with likes, people who used to react often but might be watching quietly now. Each has a different form of relationship. So I divided them into four types and placed them as a network diagram.

I placed this copy in the center of the image.

'The comment section is not a spectator seat, but a co-production seat.'

What I learned from making it

There are three things.

If you don't decide on a protagonist, the image will fall apart.
I placed the characters after deciding what the page was meant to show. When characters become the protagonists, the data melts into the background. The characters on an analysis page are fine as guides.

Don't let unverified numbers speak in images.
Use only numbers with evidence. If 'AI-generated ratios' are included in a donut chart, that image becomes a lie. Even if it looks beautiful, it's not an analysis if the numbers aren't accurate.

Structure is more interesting than ranking.
For both reader maps and mental maps, showing 'what kind of relationship' it is provides more information than 'what rank' they are. Numbers are evidence, and structure is discovery.

What I want to try next

This time, it was an analysis of my own note account.

After making it, I thought this format could also be applied to monthly analysis reports for membership members.

Of course, when analyzing other people's accounts, consideration for privacy settings and the handling of reader names is necessary. So, if I were to do it, it would be best to start with an opt-in system and deliver it as an individual HTML report to the person themselves.

I could post samples and production logs in public articles, and deliver the individual analysis content to the person themselves. That way, I might be able to continue it without strain.

Like this time, when I look back at my activities from the outside, it becomes a little easier to see 'what to try next'.

Conclusion

Analysis is not just about looking at numbers.

To make the things I've continued, the relationships that have grown, and the flow of change visible.

These five sheets have become a map for me to re-examine my note activities from the outside.

Rinka's word

……You can tell when you categorize them, right? They aren't just increasing randomly.

I think reviewing what you've built up is also a trigger for thinking about your next challenge.

If you are looking for a new challenge in the IT/Web industry, the job site Green is also an option that can support that first step.

PR | Supporting the job change site Green

👉 "Good Morning Girlfriend" is posted every morning at 7:30 at @ohayo_kanojo

#HowToContinueNote #TriedIt #NoteManagement #AIUtilization #GenerativeAI #Design #DailyNote #Learning #TrialAndError #Diary #Reflection #Record #Creation #Column #Realization #Gratitude #Life #NoteManagementLog #CreatorTrialAndError #TriedItRecord #ImageGenerationAIUtilization #ChatGPTCreativeClub #AIIllustrationAira