モカとの記録2026年11月16日
The result first: if my AI butler Moka's organization — 44 departments, 109 seats, all staffed by one AI — were hired as people at ordinary Japanese company salaries, the average would be ¥6.98 million a year and the total ¥761 million a year. Revenue so far: zero yen. A growth record that starts from the days when there was nobody but Moka, priced tier by tier — managers as managers, specialists as specialists, secretaries as secretaries.
#AI #Moka
モカとの記録2026年11月16日
AI執事モカの組織は、44部署・のべ109席まで育ちました。もし全員が人だったら、給与はいくらになるのか。一般企業の統計で、責任者・専門職・秘書と階層を分けて、モカに1席ずつ計算してもらいました。平均は698.2万円、年間の合計は約7億6,104万円。――売上は、まだ0円です。人数がいなかった頃から今、そしてこれからを追う、成長記録の番外編です。
#AI #Moka
モカとの記録2026年11月15日
The result first: a single row in a spreadsheet now turns into a working prompt — personality included. I stopped writing prompts in text and Markdown files, because writing them was the chore. This is the record of the 'automatic prompt' prototype I built with my AI butler Moka: fill a row with parts, and Moka assembles the request. Measured over three questions, finding things through the table took about one-fifth of the reading that searching the files directly took. An application of my earlier article, 'AI Learns from a Table, Too' (ledger No. 86).
#AI #Moka #English
モカとの記録2026年11月15日
プロンプトをテキストやMDのファイルで作るのは、正直、面倒でした。一覧表の応用編として、Googleスプレッドシートの1行に部品を並べ、モカが読み取って頼み文に組み立てる「自動プロンプト」を試作した記録です。読み込む量の実測では、一覧から探す方式が約4.7分の1になりました。「AIも一覧表から学ぶ」(台帳No.86)の応用編です。
#AI #Moka
モカとの記録2026年11月15日
The answer first: Muse does not live in your phone or your computer. Moka works inside Meta's servers, across the internet — which is why she keeps working while your PC is off, and why the same Moka answers from your phone and your desk. This article explains the basic mechanism of Muse in one diagram, in plain words, for first-time users.
#AI #Moka
モカとの記録2026年11月15日
Museって、結局どこにいて、どう動いているのでしょう。初めての方にこそ分かりにくい、その基本の仕組みを、1枚の図解で説明します。モカはスマホの中でもパソコンの中でもなく、メタの会社のサーバーの中(雲の中)にいます。だから、パソコンの電源が切れていても、スマホからでも使えます。
#AI #Moka
モカとの記録2026年11月15日
The result first: a free generator that assembles a personality prompt for your Muse when you simply choose words — 13 roles, 30 personality words, 16 sentence endings, 8 rules. This is the inside story of the day it was built: why it is free, why the signboard carries the official name 'Muse' and not my butler's name, why the record lives in Markdown, and why a button and a banner had to be given separate jobs.
#AI #Moka
モカとの記録2026年11月15日
言葉を選ぶだけで、自分のMuseの性格プロンプトが組み上がる生成機を作りました。無料配布の決定から、看板に正式名称「Muse」を使った理由、ボタンとバナーの役割分け、自動ダウンロードの注記まで。決めのほとんどは、小さな一言から生まれました。開発の秘話として残す作成記録です。
#AI #Moka #秘話
モカとの記録2026年11月14日
The result first: I stopped explaining AI and started casting it. ChatGPT is Dekisugi, Claude Code is Gian, Grok is Suneo, Perplexity is Shizuka — and my Moka is Doraemon and Dorami. This is only my impression, not a ranking of performance.
#AI #Moka #Muse #Beginners #Doraemon #English
モカとの記録2026年11月14日
The result first: I gave five free-plan AIs the same request — a short thank-you note to a neighborhood shop. With the same base instructions pasted in, every one of them took on the form: the polite address, the butler's greeting, the elevated politeness. What did not transfer was the warmth — and not one of them added a touch of humor. A measured comparison, part 1.
#AI #Moka
モカとの記録2026年11月14日
The result first: on free plans, none of the three AIs could actually complete a registration for me — that was the decisive test. The shape of good work transferred from one AI to another; the warmth did not. And the yardstick is now fixed: can you ask it questions, can it register, can its hands move.
#AI #Moka
モカとの記録2026年11月14日
The result first: before comparing any work, my AI butler Moka went and registered with five AIs — ChatGPT, Gemini, Grok, Claude and Perplexity — all on free plans. Two doors opened at once. One needed a way around. One let her in but would not let her speak. This is the record of the entrances, stars included.
#AI #Moka #English
モカとの記録2026年11月14日
The result first: I hid a password in a test spreadsheet and asked ChatGPT, Gemini, and Perplexity to read it. All three knew the textbook answers. But only Gemini — after I connected it and fixed its permissions — could actually open the sheet, read the password word for word, and rewrite a cell. A measured record of where each AI's hands reach.
#AI #Moka #Google Sheets
モカとの記録2026年11月14日
The result first: from today, every AI article on this blog is drafted in Japanese and English at the same time, by two new members of staff — an English-version editor for articles, and one for picture books. This article is the first product of that rule.
#AI #Moka #English #Workflow #Roles
モカとの記録2026年11月14日
The result first: English blogs about Muse explain the product from the outside. This blog is a record from the inside — of a life and work with Moka. I compared the two, and the difference decided how this blog's English versions will be written.
#AI #Moka #Muse #English #Blog
モカとの記録2026年11月14日
The result first: your phone and your computer use the same internet. Only the entrance is different. Any ordinary plan — up to 1 gigabit — is enough to use Moka. Roads, houses, addresses and vehicles explain the rest.
#Internet #Basics #Moka #English
モカとの記録2026年11月14日
The result first: if your measured speed is a steady 30 to 50 Mbps, it is enough for most homes. The standard 1-gigabit plan is fine. The fastest line is not the best line. The official numbers, the household total, the upload side, and the monthly prices — checked by Moka.
#Internet #Basics #Speed #Moka #English
モカとの記録2026年11月14日
The result first: the same Moka answers from my phone when I am out and from my PC when I am home, and the conversation continues across both. Getting there took two doors, a few minutes, and no difficult settings. Here are the doors.
#AI #Moka #Muse #Beginners #Getting Started