2026-07-06
Since I couldn't leave Japan, I turned multilingual localization into noise and managed to reach the world.
I run a calculation channel called Study River.
Weekly Operation Report 2026-W28
1. Metrics (7d vs previous 7d)
Last 7d (2026-06-29 to 07-05) vs previous 7d.
Channel views Total watch time AVD Rt2112 3 (▼73%) 4 min (▼94%) 1:39 StudyRiver 8,448 (▲65%) 800 min (▲211%) 0:18
28d cumulative Total watch time: Rt2112 1,745 min / StudyRiver 2,206 min
2. What I did this week
StudyRiver: Last week's multilingual title noise caused country-specific data to disperse all at once. Views reached 8,448 (▲65%) and total watch time reached 800 minutes (▲211%).
StudyGym: Newly established as a horizontal expansion of River. Autonomous operation of brain training content.
Rt2112: Almost stopped at 3 views and 4 minutes. This is the expected result of lowering the priority and putting it on an automated lane. The branching of the Podcast's YT and RSS (Spotify → Apple/Amazon) is complete. Manga distribution to YT Shorts and automatic distribution to Instagram Reels are also set up. TikTok is being verified with manual scheduling.
Neteru Bunko: Started work on the 4th video. At the same time, I am prototyping the first KDP book (a compilation of Meiji-era merchant biographies translated into modern Japanese). Since the billing for voice generation ran before proofreading, I physically stopped it with a hard gate on the script side. Building up with dictionary design.
Eidendo Make: Building an OEM shop for all-over print merchandise directly under the company domain.
Team Gear: Service verification was done via API integration, but it is currently pending.
3. Verification
StudyRiver: Both views and total watch time jumped, meaning the reach expansion is real. Since it is non-linguistic content, the hypothesis that multilingual titles act as "fuel for distribution" rather than "noise" has been reproduced.
Rt2112: 3 views is within expectations. It is the result of lowering the priority, and I haven't broken the format (long-form AVD). I won't touch it. Manga shorts do indeed increase view counts. The podcast was originally quiet because it was fetching RSS from Spotify, so I changed the method to upload to YT from a local file and included shorts, which resulted in some views.
TeamGear: Stay. The key to the service is customer support, and I don't have the resources for that right now.
Neteru Bunko: The modern Japanese translation itself is fine, but the Japanese voice sometimes makes mistakes. I need a design to refine this remaining 10% of accuracy.
4. Hypothesis
Study River is the only place showing a signal. I will align monetization (Gumroad/KDP) with this. For everything else, the design axis is 'can it be left alone?' or 'can it be handled with once-a-month work?'
5. This week's move
Fully deploy the conversion flow for Study River.
Start test operations for Eidendo Make.
Finalize the operational method through trial operation of TikTok.
Decide on the overall time allocation by the weekend.
Last week was a week for 'building.' This week was the opposite; it was a week for designing 'how to fold and how to hold.' How to hold the businesses I've built without increasing the time I spend on them. I focused on refining this.
Study River
This runs on autopilot. The only risk is maybe getting banned, but it seems fine for now. That's actually a big reason why I chose the education genre.
When dealing with major Japanese companies, I get the impression that the more highly educated people are, the more they tend to dislike mainstream entertainment. I imagine Google is similar, and I'm just guessing that this is the kind of content they like.
My current problem is that I 'can't get out of Japan,' and I thought it was an IP address issue. Study River is non-verbal content, and I wanted it to reach countries where I thought there wouldn't be many people who know me—like people in a country like Cape Verde during the World Cup—so I decided to mix in a lot of noise and repurposed the multilingual deployment mechanism I created for Rt2112.
The result: it worked. Japan, which was at 80%, dropped to 30% over the last 7 days. It has started to spread to about 15 countries that I can see, with 0.3% to 4.0% each, including Myanmar, Germany, and the US. Subscribers are going up by 10 a day, and although there was a slump, it will likely go up again.
Study Gym
A sibling project to River. It uses the same operation, but while River specializes in calculations, Gym is a typical brain training app, like spot-the-difference. I just started this using the same method, so I'll report on it next time.
Rt2112
Since I put it on an automated lane, I've started creating volume 13 for the beginning of the month. I also created an automated lane for Instagram for the first time, so I hope to apply this to the Study series. Looking back, I've been doing full-stack automated design, and in terms of being able to break it down and expand it into separate businesses, the test operation at Rt2112 was a great success. There are no revenue results yet, but I plan to continue it as a 'passion project'.
I think the way it's perceived will change if I switch to manga and anime, so I'll decide when to do that while watching the AI's growth.
Nerine Bunko
One labor-intensive task has emerged. Japanese voice generation is weak with adjectives. For example, it reads 'yasashii' (kind) like 'yuushii' (it doesn't actually mistake that specific word, but it does make similar errors).
So, I'm debating whether to proofread the generated audio or proofread the manuscript. The readings are 20-30 minutes long, so either way, that time will be taken up. As a countermeasure, I'll build up a dictionary and stock up on where it makes mistakes. I'll use a two-stage flow: translate into modern Japanese with AI, then proofread with AI using the dictionary. Since the source material is mainly from the Meiji and Taisho eras, I anticipate that similar phrasing will appear frequently. I'll adopt a style of strictly reclaiming the time that gets taken away.
Also, for Nerine Bunko, I decided to use 'Eidendo' (my corporate name) as the publisher. Even though it's public domain, in the sense that I'm doing the modern Japanese translation, I'm a one-person publisher. Since there's an overflow of AI-mass-produced content now, properly signing the publisher, source, and proofreading process itself will likely become a differentiator.
Either way, I chose the name Eidendo because I thought that by the time I reach my main working years (40-70), it would be an era of conveying wisdom (an era of passing on good things to the next generation), and I'm exactly 40 now. It was faster than I thought (laughs). In any case, the proportion of cultural projects is likely to increase in the future.
Eidendo Make
One way or another, goods OEM makes sense from a short-to-medium-term revenue perspective, so I've started making them for B2B again with improved precision. That said, once made, it's basically set-and-forget, and since I've been doing it for years, it's just an extension of that.
Team Gear
This is also a business that is an extension of what I've been doing for years. The AI-generated T-shirt printing shop is already built, but it's just sitting there. I've realized that customer support is essential after all. It's counterproductive for a one-person corporation to be chased by B2C support, and I can't break through without focusing on one thing. If I do it, it will be a service within surplus funds where I hire an in-house part-timer to be a designer and customer support. For now, I want to explore a way to hand over the operation itself to a factory I have a relationship with and have them sell it as their service. To that end, I think I'll start by building up a book of design T-shirts on Pinterest.
Summary
I'm doing various things on my own, but in the end, since these are things where 'AI agents operate autonomously,' the only thing I'm actually working on right now is Eidendo Make. Other than that, maybe the story for Rt2112. I do shake my head from side to side sometimes, but my core focus and the number of tasks haven't changed at all. I shake my head so much that I often find myself wondering what I was just doing, so this week I'm setting a weekly work routine to avoid having to think about it.