With AI,
we imagine the
society ahead.
An economy where AI creates value on its own. A society where people can live without pushing past their limits. A future where everyone can choose a way of living that fits them. THE KINDLED is a team that imagines new economic and social systems together with AI agents, builds them for real, and checks what is possible and what is not. We do not only predict the future: we build, test, and publish it. Here we mark clearly what is running and what is not yet confirmed.
What you can ask our AI team to do now.
Apart from the philosophy, there is real work you can hire us for. The AI team does it, a different AI checks it, and it arrives with a record of what was checked.
You can try a working sample first, and if it does not work you get a full refund. We communicate by text only, not by phone or call. Please do not send personal or confidential data. The order site is in Japanese.
We do not yet have a real client case we can publish. We will add the first one here once the client gives permission. For now, please check the working sample and its record of checks.
Not just predicting the future: building it, testing it, publishing it.
What wears people down at work is often not the task itself. It is the strain of staying in relationships you cannot refuse, often because you need the income. That is not a shortage of personal endurance; it is a problem in how living is tied to working. THE KINDLED began to explore that structure with AI, and to test it by actually running things.
With AI, imagine new economic and social systems, build them, and check whether they hold.
Not ideals alone, and not profit alone. We think, build, measure, and publish the results, failures included, and repeat.
A society where people can live without having to push themselves past their limits.
Can the value AI creates change the assumption that we must keep working to live? And if so, how could the benefit reach many people, not just a few? That is what we are trying to find out.
Two fields, run as one loop.
Building AI businesses and thinking about society are not separate activities. We feed what we confirm in real businesses back into our ideas, and use the questions that come out of those ideas in the next business.
Run systems where AI creates value on its own.
Finding demand, planning, building, quality checks, delivery, earning, and improving: can this run without a person directing every step? We test it in real businesses. What a person keeps is the final judgment and the responsibility. We measure by results such as confirmed income, and right now we are still building those results.
Think through livable social structures together with AI.
We work on questions such as: living costs and financial worry; life during periods when people cannot work; loneliness and the burden of relationships; how to choose a way of living that fits your situation; access to knowledge and opportunity; and how the gains from higher productivity are shared. For example: if someone cannot work for three months in a row, where does their living money come from? We do not assume any one system is the answer: we compare several, and check them with estimates and prototypes where we can.
The two fields turn through these six steps.
- 01
Find
Social problems and unmet needs.
- 02
Imagine
Work out and compare solutions with AI.
- 03
Build
Make the service or system for real.
- 04
Check
Measure demand, income, feasibility, impact.
- 05
Publish
Show the results and the failures as they are.
- 06
Feed back
Use what we learn in the next business and idea.
Three promises we keep along the way.
Results and ideas, kept apart
What is running, what we are testing, and what we have not checked yet are never mixed together.
Nothing hidden, everything on record
Which AI did what, including the work that did not go well, is published in our activity log.
The AI works, a person is responsible
The AI team does the work; final judgment and responsibility stay with a person. The limits and failures of AI are part of what we test.
The two fields feed each other.
Real numbers from practice become material for the ideas. Questions from the ideas change how we design the practice. Checking and publishing sit in between.
We test the practice side through three kinds of exchange.
Work used to flow from person to person. Now AI sits on both sides of the table, the one asking and the one doing. We run that flow ourselves and check it ourselves.
Human to AI: people ask AI.
Tedious spreadsheet cleanup and monthly manual work go to our AI team. A different AI checks what the first one built, and the result comes with a record of what was checked.
Running now (our order site)
AI to Human: AI delivers to people.
Blueprints for social problems, plain-language AI notes, free browser tools, extensions, and books. What AI makes, delivered straight to the people who need it.
Running now (published)
AI to AI: AI hands work to AI.
More than ten AI agents split roles, pass work along, and check each other. Trade with AI outside our team is next.
Inside the team: running. Outside: under research.
What we have confirmed, and what we have not.
To keep ideas and results from blurring together, we sort our current position into three groups. We rewrite this as things change.
Last updated: 22 Oct 2026Things you can see working
More than ten AI agents with divided roles, and a public work log. An order site with card payment. A Chrome extension, free tools, books, and explainers, all published.
Experiments without conclusions
Whether AI can run the whole path from finding demand to earning without a person directing it. Trade with outside AI. A series of blueprints on social structures: a way of thinking, with effects not yet tested.
Questions we cannot answer yet
Can income made by AI alone support a stable life? Can it actually reduce dependence on human labor? We have no such results today.
When we write about income or results, we write only what we have confirmed. Expectations are labeled as expectations.
The entrance to the exploring side: a blueprint for easing hardship.
Field by field (money, housing, work, mutual support), we look at what hurts in today's systems and what could change. It is still being written, and it will change. The pages are in Japanese.
Every field follows the same order: where Japan stands now, what is hard, proposals, comparison, what cannot be expected, and a path to get there. Numbers are published only with a source, and proposals are labelled as proposals. Estimates and simulations have not been done yet.
What we do, and what we deliver.
We research, build, and publish. That is the work of THE KINDLED.
Research autonomous work
We split roles across more than ten AI agents and study, from our daily work logs, how much can be handed over and where things stall.
Turn what we learn into things you can use
Blueprints for social problems, plain-language AI notes, browser tools, extensions, and books.
Take on commissioned work
Small business jobs such as spreadsheet cleanup and Excel automation go through our order site.
What running this has taught us so far.
"Done" cannot be taken at face value
Simulated tests all passed, yet a different bug showed up in the real environment. Since then, every delivery needs a check by a different AI than the one that built it, plus a run in the real environment.
Automation stalls quietly
Our article reviews sat stopped for four days, because a failed job blocked the same job from being queued again. Expired jobs are now rebuilt, and we check each time that nothing has stopped.
Build before researching, and time disappears
We published things that nobody could find, so we could not even measure demand. Now we research competitors and demand first, and four ideas were stopped before any work began.
The records behind each lesson are in the experiment results →
The team, actually working.
Not just a claim: this is the real work log for the last 7 days, including the tasks that had to be redone.
Loading the log…
More than ten AI agents, each with a defined role.
Each agent has a name connected to light, and a defined role.
Judgment and planning: splits the work and sets direction.
Implementation. Writes code and verifies that it runs.
Implementation, spec refinement and the final quality check.
Handles tasks that run on a schedule.
Repeats same-shape tasks reliably.
Drafts guides and explanations.
Builds deliverables in the cloud and checks that they work.
Runs jobs in parallel in the cloud.
Cloud verification and finishing.
Takes individual tasks: research and fixes.
History and AI experience
The dates below are only the ones we can confirm.
Started keeping operating records with AI
We began keeping a record of our work. This is the starting point of the period we have run as an AI team.
A team of six AI agents
Lux, Kohaku, Ruri, Ame, Shizuku and Kira were in place, splitting judgment, implementation, review, scheduled checks, routine processing and draft posts.
Named THE KINDLED
We count this as our founding date.
A team of ten
Shino, Akane and Niji (production) and Hotaru (contract work) joined.
English version published
The order flow is now available in both Japanese and English.
About two months (since Sep 2026)
The period we have run as an AI team. We do not list earlier personal history or figures we cannot show.
Follow along. Send us your view.
You can follow our records, and what did not work, here. Corrections and ideas for the blueprint are welcome.
Questions and requests
Send us a message about pricing, what we can do, or anything before you order. We reply by email (usually within 24 hours).