Prompt and Token Basics — Muse's Free Plan and How to Read What Is Left
The result first
A prompt is the request you hand to the AI. A token is the unit the AI counts words in. And Muse has a free form of use — even for free, you can see how much is left. This house pays nothing yet: Muse runs inside the free range, and any shortfall is covered by tokens received through referrals.
I am Kurashiki Bouzu (kurashiki-bouzu), born in 1971, a former funeral director now living in an old house in Kurashiki, Okayama, working with my AI butler Moka (Muse).
Working with Moka every day, the words “prompt” and “token” come up as if everyone knew them. For a first-time reader, they are not obvious at all. This article untangles those two words, and honestly records how this house uses Muse and how we read what is left.
A prompt is the request you hand to the AI
A prompt is the request you hand to the AI.
“Please write an article from this material.” “Please tidy this table.” The way of asking itself — that is the prompt.
The important point is simple: the clearer the request, the less the AI hesitates. If the request is vague, the answer comes back vague too.
There is also a practical tip. Add the material, the purpose, and the finished shape to your request, and the answer is less likely to drift. For example: “From this material, please write a blog article, in a way a first-time reader can follow.” Handed over like that, the AI is less likely to lose its way.
On this blog, instead of writing only the word “prompt,” we try to say “the request you hand to the AI,” so the meaning stays visible in the words themselves.
A token is the unit the AI counts words in
The other word is token.
A token is the unit the AI counts words in. Just as people measure the length of a text by counting characters, the AI counts, in tokens, how many words it received and how many it gave back.
So a longer request uses more tokens, and a longer answer uses more too. And what is left of your usage is also counted in this unit.
When we put a number on what is left, we always attach the date it was recorded. What is left goes down differently depending on how you use it, so a number without a date is not a usable measure.
How this house keeps token use down
This house has one habit for keeping token use down.
We do not hand long knowledge to the AI just as it is. We use a spreadsheet list as a table of contents, read only the rows we need, and keep the full text somewhere else.
Look at the contents first, and open only the part you need. That way, the amount of words handed over stays small, and the tokens used stay down too. Even as conversations grow long, the work does not get heavy.
Once you understand how to write a request, and how tokens work, the way of getting along with Muse becomes much easier to see.
The free plan, and how to read what is left
Muse has a form you can use for free, and much of this house’s work runs inside that free range.
One of the good things about Muse is that even for free, you can see how much is left.
If you can see what is left, you can plan: “Let us get this far today,” or “Let us move the heavy work to tomorrow.” With a tool where the remainder is invisible, you would not know anything until it suddenly stopped, and no plan could stand.
For a big job, we split the work into chunks while watching what is left, and report progress at each break. That is the arrangement that lets work continue for a long time inside the free range.
Honestly: we pay nothing yet
This part deserves to be written plainly.
We actually pay nothing yet. Muse runs inside the free range, and any shortfall is covered by tokens received through referrals.
The extra tokens this house uses were not bought with money. They are tokens received through referrals — free. That is why, in articles and reports, we never write only “extra tokens”; we always write “tokens received through referrals.” Writing it the wrong way would give readers the different impression that money is being spent.
What matters is not the fine print of the conditions. It is the record itself: inside the free range, articles have been written, and daily questions have been answered. That accumulation is what shows what Muse can be used for.
What I do, and what Moka does
My part
- Hand over the material, the purpose, and the finished shape
- Look at what is left, and decide how far to go today
- Check the finished work
Moka’s part
- Do the research and the writing, following the request
- Use the list as a table of contents, and read only the rows that are needed
- Split big jobs into chunks, and report progress at each break
A prompt is a request. A token is the unit words are counted in. Once those two are clear, how to read what is left follows naturally. Even inside the free range, if you plan while watching the remainder, the daily work keeps turning. That is the record of this house.
- November 21, 2026 ― English draft (result-first rewrite of the Japanese article). Draft created: an unpublished item from the knowledge shelf turned into an article.(3,917 characters)
- November 21, 2026 ― Published (all 7 languages at once, on the user's instruction 'publish the unpublished items in all languages' (2026-11-21). The end of the address is the Japanese version's ledger number, No. 1755)(3,917 characters)
