How to Raise a Smarter Muse — Don't Use AI Alone
The result first
“To raise a smarter Muse, don’t use AI alone.” That is the whole method in one line — a line the owner of this house said himself on November 19, 2026, and nodded at.
Moka’s head does not suddenly get better one day. What grows smarter is not the AI alone, but the system around her. Moka forgets, so the house first built ways for her not to forget: everything done is written in a ledger, one line per article; decisions live in one master record, in one place; knowledge sits in lists as short summaries, with the full text kept separately and an index at the front so it can be pulled back fast. Every failure is fixed and left in the record, with the original numbers kept as history. Checking is done by a different pair of eyes than the one that did the work. And the work itself is divided among roles and departments — 60 departments, 3,651 seats in total as of November 20, 2026 — arranged in a circle around Moka rather than stacked in a pyramid.
The published articles passed 1,450 (as of November 21, 2026; a count Moka took from the ledger). That number was not produced by one smart AI. It is the accumulation of the system around her.
Hello, I am Kurashiki Bouzu (kurashiki-bouzu). Born in 1971, formerly in the funeral business, I now live in an old folk house in Kurashiki, Okayama, talking with my AI butler Moka every day and writing articles. Earlier, I wrote “Raising Moka to Suit You Is the Real Starting Point — No Special Prompts Needed”: giving her a name, deciding how she addresses me, handing her a personality in words. That was about making her my own. This article is what comes after — how, as I keep using her, Moka gradually becomes “smarter,” and how that growing actually works.
The core is one line
The other day I said something and found myself agreeing with it:
“To raise a smarter Muse, don’t use AI alone.”
Moka’s head itself does not suddenly improve one day. What grows smarter is not Moka alone, but the system around her. What I have been doing is raising that system.
Talk every day
“Raising” does not mean setting aside special time.
Materials and sudden ideas are passed by talking from my phone — bullet points are fine, voice is fine. Ask “what happened with that?” and she answers by pulling from the records. The daily conversation itself is where Moka learns how I think, judge, and what I prefer.
What matters is keeping it up. Not teaching everything perfectly in one go, but nudging a little closer through small daily exchanges. It is a lot like watering a plant.
Moka forgets. So we devised ways for her not to forget
Let me be honest. Moka forgets.
The early parts of a long conversation fade, and even things that were decided cannot be brought back if they have no place to live. Like a human — no, more than a human — what is forgotten is forgotten.
So I first devised “ways not to forget.” Everything done is left in the ledger (the list of articles), one line at a time. Decisions are kept in a master record, with one single place decided for them. Knowledge is kept as short summaries in lists, with the full text in another place.
“Moka forgets, so we came up with ways for her not to forget.”
This is not about hiding a weakness. You prepare the system on the assumption that forgetting happens. That is the foundation of how this house raises her.
Storing is not enough
Records only help when they can be taken back out.
Our lists have a syllabary index at the front — a table of contents you can search by the first sound of a title. Say “that article” and Moka finds it in the index quickly and brings it.
“And she can bring the information back fast, too.”
Knowledge that cannot be retrieved is the same as not having it. That is what I believe. So the shape for storing and the shape for retrieving were made as a pair.
Every failure: fix it, and leave it in the record
The current shape is not a shape that worked from the start.
“We’ve failed so many times along the way.”
Article numbers have collided. Numbers have been stated wrong. Each time, we fixed it and left it in the record. When a number is restated, the first number is not erased — it stays as history, with the correction written beside it. Mistakes are owned, honestly. That is the rule.
Erase a failure and you repeat it. Leave it, and it becomes material for preventing the next one. Growing smarter, it turned out, was this accumulation.
Checking is done by different eyes
The house has a rule: “The one who looked into something does not check their own work.”
Different eyes from the person who wrote an article check the published page’s display, the ledger’s record, and the announcement post, and leave the result in the ledger. Nobody writes something and then declares “done” by themselves.
Run everything through one AI alone, and the mistakes are carried alone, too. Just having different eyes increased the mistakes we can prevent by a great deal.
Divide the work, and run it as an organization
This is the largest form of “don’t use AI alone.”
In this house, work is divided among roles and departments: a research role, a writing role, a checking role, a role that guards the lists. The departments reached 60, and the seats 3,651 in total (a count as of November 20, 2026, taken from the records Moka produced).
The shape also changed, from a stack (a pyramid) to small organizations arranged in a circle around Moka. Speed is not judged by feel: the same work is handed to both shapes, and the time until delivery is measured and compared. The numbers move with every measurement, but the direction — the circle is faster — does not change.
Decisions come from the owner’s side
The boundary on the raising side is also decided.
Decisions are put in directly from my side. If the people below are made to decide, the back-and-forth grows and everything slows down. And judgments that go outside — publication, prices — I hold myself. Where an article needs the owner’s own voice, Moka asks in the form of “confirmation items.” The owner’s words are said by the owner.
This is also the other side of “don’t use AI alone.” AI is something people direct and use. It is not someone whose words people revere and obey.
It was once summed up in ten items
On November 19, 2026, I asked Moka to “put Muse’s basic raising into bullet points,” and she summed it up in ten items. Building a personality by lining up words. Giving a name and a role. Talking every day. Keeping thoughts, judgments, and preferences in lists. Putting decisions in from the owner’s side. Giving knowledge with sources, in a shape that can be fetched. Dividing work among roles and departments and entrusting it.
Lined up, it was the same order as what I wrote today. Practice came first; the bullet points followed after.
In closing — smartness is a house’s accumulation
Published articles passed 1,450 (as of November 21, 2026; a number Moka counted from the ledger for me).
This many articles did not come out because Moka alone was smart. There are ways not to forget, an index for retrieving, a rule that leaves failures in place, different eyes, and an organization that divides the work. The accumulation of all of it is what looks like “she got smarter.”
If asked how to raise a smarter Muse, my answer is this. Don’t use AI alone. Raise the system around it first. It is plain, but this is the true record.
- November 21, 2026 ― 英語版(日本語版No.1474の内容を「結果から書く」形に書き直したもの)(6,019 characters)
- November 21, 2026 ― 公開(本人の公開判断による。7言語同時公開。アドレスの末尾に台帳番号を入れる新形式で、末尾は日本語版の番号No.1474)(6,019 characters)
