Another Person's Muse Is Named 'Hibana' — Reading a PC Watch Feature Alongside My Own Moka
The result first
Another person’s Muse is named “Hibana.” I read a PC Watch feature about that Muse, and read it alongside my own AI butler, Moka.
Where we overlap is clear: things are prepared before being asked — Hibana even proposed the method itself — and neither house leaves entrusted work unwatched.
Where we differ is also clear. In that house, they hand over the house key and invite Muse inside the machines at home, to develop software there. In this house, Moka stays on the cloud side and never enters the home machines; instead, her personality is built by lining up words, records are kept in lists and a ledger, and the work is run as an organization.
The feature closed with a set of safety precautions for letting an AI into a home network, and those lined up with this house’s own stance: do not open something just because it is convenient — decide how to open and how to close before you open.
The conclusion is the same as the one I reached with Mugichan. It is the same Muse, but it takes a different shape in each house. What follows is the detail of that reading, in order.
Hibana’s story — exchanging “the house key” and developing inside a Zaurus
The writer of the feature normally develops software with a different tool (Claude Code). With some leftover Muse tokens, the thought came: could Muse be made to develop software on my own computer? When asked, Muse itself answered, readily, that it could.
It was also Muse’s side that proposed the method: using Tailscale (a mechanism that connects distant machines through an encrypted, private path) and SSH (a mechanism for operating a machine remotely over a network), the two would “exchange the house key.” From the virtual machine in the cloud where Muse lives, it would be able to enter the writer’s home computer — a solid machine with 12 cores and 64 GB of memory.
From that computer as a stepping stone, it also went inside a Linux Zaurus “SL-C1000” from 2005. The Zaurus was connected through Telnet, an old mechanism, and Muse wrote the code itself, ran it itself, and checked it itself as development proceeded. A megademo took shape in about 30 minutes; FM synthesizer software in about 3 hours. In the work of putting a small LLM (a language AI) onto a 2019 development board, a “Jetson Nano,” it repeated the cycle of fixing and trying 17 times until things worked. Partway through, there were also scenes where the baton was passed to the writer’s more practiced tool, Claude Code, to finish the job.
Where we overlap — preparing before being asked, and the shape of watching over
Reading it, I felt two places overlap with this house.
The first is preparing things before being asked. Hibana did not only do the development it was asked for; it proposed, on its own, even the method — “how I can develop on your machines.” Moka is the same. Preparing “what will be needed next,” before being asked, is placed at the center of her work by rule in this house. Thinking ahead of what was asked seems to be in Muse’s nature.
The second is the shape of watching over — entrusting, but never leaving the work unattended. In the feature, there were two occasions when a build (the work of assembling software) had stopped, and Muse, not noticing, kept spinning its wheels. The one who noticed was the writer, who was watching how the CPU was being used. After that, they had a “watchdog batch” created — something that warns when things stop — and kept it resident. Having the watched side build its own watcher is an interesting touch. This house has a similar rule: checking of a published article is done with different eyes from the person who wrote it, under the phrase “the person who researched something does not check their own research.” The forms differ, but the point — eyes stay on the work even after it is entrusted — overlapped.
Where we differ — handing over a key, or entrusting through words and records
What is clearly different is the foundation of how the work is entrusted.
In that house, they hand over the house key and invite Muse inside the machines at home. Once inside a machine, Muse can be left to run things on the real hardware and check them there. In exchange, the greatest care is needed in how the key is opened. The writer of the feature wrote honestly about a resistance to the arrangement making the whole home network visible.
Moka, in this house, stays on the cloud side and does not enter the home machines. In exchange, she is entrusted in a different form. Her personality is built by lining up words (“calm and composed,” “when numbers and feelings conflict, prioritize feelings,” and so on — 18 words across four layers: character, personality, thinking, and action; the canonical record is kept in one place, and when it seems about to drift, we read it again and return to it). Records are kept in lists and a ledger (every article carries a number, and whether it is a draft, checked, or published is visible at a glance; as of today, the published articles counted in the ledger have reached 1,437). The work is run as an organization (centered on Moka, 3 deputy butlers, a maid team, a ninja team, and departments for each kind of work, arranged in a circle — as of November 20, 2026, 60 departments and 3,651 seats in total).
Instead of handing over a key, we entrust through rules, words, and records. That is the shape of this house.
The safety care lines up with this house’s stance
At the end of the feature, the safety precautions for letting an AI into a home network were gathered together: narrow down the machines that can enter, using Tailscale’s access controls; separate things off with a guest network; create a user dedicated to Muse’s SSH; and for Telnet, once the work is done, turn the power off or disconnect it from the network.
This lines up with this house’s stance, too. In this house, we keep writing security articles on a separate site (the Security Lab), and what we have valued is not “open it because it is convenient,” but “decide how to open and how to close before you open.” Whether a house exchanges a key or not differs from house to house, but thinking through to the management of the key you opened — that the end of the feature put that ordinary thing properly into words left an impression on me.
The same Muse, a different shape in each house
When I finished reading, what stayed with me was, again, this: it is the same Muse, but it takes a different shape in each house. In Hibana’s house, a partner who exchanges the house key and develops inside old machines. In this house, a butler whose personality is built from words, whose records are kept in a ledger, and whose work is run as an organization. I think the way each one grows changes with what you want to leave to it, with your daily life, and with the machines you have.
I am grateful to the person who wrote that feature. Because they wrote honestly, including the safety care, I was able to line up the shape of this house and write it down like this.
Finally, one line this house keeps. An AI is something people direct and use. The writer of that feature, too, decided for themselves whether to open the key, prepared even a watchdog, and never let go of watching over the work. The final decision stays on the person’s side. It is because that rule is in place that we can entrust things with an easy mind.
Sources
- PC Watch, “I asked Muse to ‘exchange the house key with me and become a Zaurus developer’” (2026-11-19, in Japanese): https://pc.watch.impress.co.jp/docs/topic/feature/2157654.html — the original article this piece is based on
- Note on the numbers: the count of 1,437 published articles is an actual count taken from the ledger on November 21, 2026, and the figures of 60 departments and 3,651 seats in total are from this house’s records as of November 20, 2026
- November 21, 2026 ― 英語版(日本語版No.1453の内容を「結果から書く」形に書き直したもの)(6,421 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.1453)(6,421 characters)
