I'm Ikehaya.
Are you working steadily today?
I saw a post by Mr. Kajitani on X.
https://x.com/kajikent/status/2093848882367344803
This is exactly right.
However, this story probably won't make sense to people who aren't using AI properly.
Today, I'll explain with examples I feel every day.
=====
As a premise, the intelligence of the AI model itself is the same no matter who uses it.
I use Fable 5, and the "brain" of Fable 5 itself is the same for both you and me. That's obvious.
However, the "quality of output" between my Fable and your Fable is probably completely different.
When making a website or an app, I think my Fable is faster and produces higher quality work.
The reason for this is that we have "trial-and-error assets." Since July, I've been running 12 Fable accounts, making everything from games to web services.
The "landmines" I stepped on during that process and the history of overcoming them are all accumulated within the environment.
Even with the same brain, what you can do changes completely depending on whether you have these "assets."
=====
Design is an easy-to-understand example.
The game I'm currently making, "Tsukimi Occulta," had such a tacky design for the first two weeks of development that it felt uncomfortable every time I touched it.
So, once the features were ready, I spent about three days thoroughly fixing the design.
I've even given this design a name: "Habutae Design." Like the smooth silk fabric, Habutae.
When you have assets like this, the next time I make a site, I can just say "Do it in Habutae style," and it will instantly produce the design I once seriously refined.
In fact, I'm currently building a new marketplace, and it's an arrangement of the Habutae design I polished through the game.
Conversely, if someone who just started using Fable today asks it to "make a nice-looking site," unfortunately, something mediocre will come out.
Even with Fable 5, the out-of-the-box design is honestly not good.
Since we've overcome that hell daily, our Fable "knows" design.
=====
The same thing is happening in game and app development.
"Tsukimi Occulta" is made with "Flutter," which has its quirks—there are many points where it bugs out unexpectedly if you miss an implementation detail.
We've stepped on all of them.
And we record every landmine we step on, creating a "Flutter Landmine Collection."
Thanks to this, I've been making most small apps with Flutter lately, and strange display errors hardly ever appear.
When I ask it to "make an app like this," it implements it while smoothly avoiding the landmines we stepped on in the past.
We've also stepped on many errors and mistakes regarding the application process, so store review rejections have decreased significantly!
====
It's the same for security.
Yesterday, there was a trouble where the email server went down due to a massive bot registration attack on the membership registration for Tsukimi Occulta.
Since we are distributing NFTs, bots are targeting them.
By this morning, we implemented solid countermeasures, and bots basically can't register anymore. This response history is also knowledge.
The next time we implement an NFT-related service, we can just say "Put in the same countermeasures as last time."
(Please try registering!)
https://vibe.co.jp/luna-occulta/tsukimidai
=====
In other words, AI intelligence depends more on
"The knowledge accumulated in your environment"
than on the intelligence of the model itself.
This is similar to a company.
When you do various tasks, knowledge accumulates, is passed on, and becomes the company's business assets.
That is happening inside a single computer.
Our AI employees understand and act on things like "Let's avoid this because we got stuck here last time" or "This is the most efficient way." They are truly smart.
That's why there is value here.
The knowledge about game production accumulated by our Fable cost a lot to build, so we will not make it public.
If we were to provide it, it would be in the form of B2B or consulting contracts.
It might sound stingy, but this isn't something to give out for free. Please understand it as a trade secret.
=====
So, the environments of people who are aggressively using AI at the forefront have accumulated overwhelming knowledge, allowing them to advance their businesses at explosive speeds.
Our environment has also evolved tremendously in this past month. September will likely be even more incredible.
I think there are still very few people who can keep up to this point.
So, a final announcement.
Yuki-maru will be holding a "free seminar" on Claude Code.
https://x.com/Yuki0_0Maru/status/2094171291846955120
He has experience as a training instructor and is a caring person, so he is the right fit for beginners.
I guarantee there are no expensive back-end offers.
If you heard today's story and thought "I don't understand this at all," please go learn from Yuki-maru.
August ends today. It was a great month.
Let's keep building steadily.
=====
I deliver daily columns on Substack. Check it out too!
[Ikehaya's Substack is here]
