Jun Sasaki / Log

2026-08-03

Clear contrast in performance

Currently practicing AI autonomous operation for YouTube and various SNS.

Weekly Operation Report 2026-08-02

Target Period: 2026-07-27 to 2026-08-02

Metrics

Weekly Metrics 2026-08-02

Last 7d 2026-07-27 to 2026-08-02 vs Previous 7d 2026-07-20 to 2026-07-26

Channel views Total watch time (min) AVD
Rt2112 2 ▼78% 1 ▼90% 0:33
StudyRiver 13,467 ▼37% 1,025 ▼44% 0:20
StudyGym — Error —
StudyJaEn 11,601 ▼32% 616 ▼34% 0:14
MazeMarathon 17,164 ▲17% 3,177 ▲0% 0:33

Reference (28d cumulative)

  • Rt2112 28d cumulative total watch time: 87 minutes

  • StudyRiver 28d cumulative total watch time: 7,203 minutes

  • StudyGym: Acquisition failed — The read operation timed out

  • StudyJaEn 28d cumulative total watch time: 2,486 minutes

  • MazeMarathon 28d cumulative total watch time: 6,598 minutes

What I did this week

  • YouTube: Since the operation method for scheduled posts was inconsistent, I turned the operation of the stable channel into a skill and deployed it to other channels. It should be fine as long as the PC doesn't crash. The viewership ratio from Japan for Study River jumped from 36.6% to 98.7%, and overseas views dropped by over 90%. However, total views remained flat, and audience retention did not drop. Setting aside Lang-jaen, Maze is showing strong growth.

  • Shoten.me: Fully automated store operations. Running 1 store and shelf addition per day via Cowork. Short videos are distributed to Instagram and YouTube once a day. During the process, I identified and fixed a bug in the affiliate funnel and added verification. The limit on the number of shelves was also resolved.

  • Common issue with short-form distribution: Only TikTok posting is not automated. The current method seems to be against official terms of service. Currently, I am operating it with brute-force scheduling, but that definitely won't last, so I need to figure out what to do.

  • Eidendo: The company website is being updated daily by Codex. I will review it when I have time and provide update instructions.

Verification

  • YouTube Shorts are still stuck in Japan. They went overseas but came back. (This is troublesome). My view is that it was pulled in by senior audience retention, resulting in 90% Japan. Since it might turn back again, I will maintain it with hopes for the 0.5% in Serbia, but in the worst case, I will consider removing Japanese from the multi-language options.

  • Shoten.me is starting to behave like the movements of a real store clerk. There are two current issues: I can 'add' items to the shelf via MCP, but I cannot 'remove' (delete) them, and I cannot rearrange the books, so I will implement these as soon as fable is released.

  • Despite the hassle of TikTok, I decided to shift my perspective and just build a common marketing SaaS tool instead. It seems applicable to all channels. Even though it's a SaaS, it's just for my own use.

Hypothesis

A slight digression. I recently saw something like a fully automated routine for YouTuber content planning, but I wouldn't recommend that. Since even channels that are growing now are leaning toward functional content, I feel that automating personality-driven content is a bad move. However, the implementation itself is possible.

Next week's move

Continuation and improvement. Also, I will try implementing the marketing tool.

This week's focus

Full-scale operation of Shoten.me has begun.
Currently, new store shelves are updated at 7:00 AM, and existing store shelves are updated at 7:00 PM.
Even though they are called stores, it's like a bookshelf corner in a bookstore, so it's the same feeling as a bookstore clerk restocking books.
Automation of 'Oh, I bet this will sell.'

A fundamental point.
When building a service, I operate on the premise of how crudely I can cover the 1 to 10 phase, so the policy of automating the stacking of shelves and books itself was in place from the design stage.
Also, although it is a curation SaaS, I assume no one will use it for a year, so I am also thinking on the premise that I will be the one adding the shelves.
By doing this,
the state of 'I tried making it, but no one uses it' is eliminated, and I make it lively myself.
If people don't come, then it just becomes a service where I run a bookstore, so that's fine.

Services themselves can be easily created with AI, and aside from idols who have fans from the start, since there is no one there, I recommend building with this mindset for those people.

There is no one for you!! Sad

Overseas disappeared and Japan remained

About YouTube.
Even though the content called Study River was distributed worldwide with multilingual titles, it ended up being a joke where 94% of the audience was from Japan in the last 7 days.
This was quite unexpected. The number of views dropped, but seeing that the audience retention rate hasn't changed, I think 38% are seniors, and most of them are in Japan.

As a countermeasure, the only thing to do is to reduce the number of videos from 8 to 6 per day and wait for them to be distributed overseas.
Regarding this, there is nothing I can do on my end, so I wonder what to do, and it's a difficult situation because Japanese people don't tend to subscribe to this type of entertainment-based educational content, nor do they leave comments.
Conversely, the English composition channel is seeing a massive increase in subscribers, and since this is problem-solving educational content, there are many Japanese people who subscribe just in case. (I say 'just in case' because the total number of views hasn't increased that much.)

What's scary about this hypothesis is the Maze channel.
It's growing extremely well, and while the long-form content is mainly for Southeast and West Asia, 95% of the shorts are from Japan. Recently, senior comments have started coming in, and I'm worried about being pulled in that direction.
The senior ratio is low, so I think it's fine, but...

In other words, it means I should make content for Japanese seniors.
I am making non-linguistic cross-border content to escape that, but a considerable dilemma continues.

TikTok diss

Man, YouTube's open nature is the best. It can be automated immediately. There are targets like bans, but the entry point is wide. Instagram is also a hassle because it requires a server, but since I have three running now, if you don't know how, I think it's fine to leave it to Fable. Claude in Chrome will do it for you.
Now, TikTok. You are terrible. Telling me to capture this, make a site, make it look like a brand, and so on, and in the end, it's a no-go. It feels very Chinese (diss).

If I were to create one loophole, it's the common third-party apps that allow you to operate TikTok. Apparently, those are okay. (I don't understand the logic, and even if they say it's okay, there are often patterns where it becomes a no-go.)
The fact that it's a hassle to enter means that if you break through, there's money to be made, so I think I'll just accept that and try it.
This is because the persona users of TikTok are inseparable from the next business I'm launching.
However, one thing.
Damn it!

Summary

I've cleared things up in a word, so I'll keep clicking away while watching Koshien this week too.
Yesterday, an idol buzzed about making a streaming service with Claude Code, and I think that's fine for creation in the AI era.
Purpose: What do I want to do?
Means: Ask Fable.
This alone takes shape.

I think you should just keep doing it, and if you increase the metabolism of the economy like that, I think the situation where Japan is in trouble now will disappear, so it's all about taking on challenges.

Also, it doesn't matter, but I recommend dumping logs on Monday morning.
Because Claude's credits run out and I have nothing to do (bitter smile)

Advertisement