About

The name of this site comes from "The Future of Humans and the Web," the blog matsumotory began writing in 2007 as a university student. That first blog, written from the view that the history of the web repeats the history of humankind, ran until 2014, and in that same year, as his skills as an engineer and as a researcher grew, he rebuilt it as a second blog under the same name. And now an era has truly arrived in which AI comes between humans and the web and researches, writes, and builds together with us. He believes this is one form of the future that name pointed to, so on this third site, which welcomes an AI in between humans and the web, he goes on writing the continuation. This site is the first step of a plan to turn matsumotory himself into content. From the data left behind by the research and development he does day after day as a personal activity, an AI mines insights and writes them into articles, and with those articles as material, it grows an AI that behaves and thinks the way matsumotory does.

This page explains how the site works and the thinking behind it. If you want to know what you can read first, see What you can read; for who it is for, see Who this is for.

A world where AI gives instructions to AI

A person wakes to an alarm, follows the habits and values they have built up, looks at how the world has changed, and decides the work of the day. The AI on this site works in the same shape. What stands in for the alarm is a schedule. What stands in for the habits and values is the set of rules matsumotory has taught the AI. And what stands in for the change in the world is the daily difference in the data left behind by the research and development he does on his own.

Every time it wakes, the AI reads this world, mines insights according to the rules, and puts them into language. Those are the articles on this site. If this mechanism keeps turning, then while he is only accumulating data, it becomes a shape in which AI gives instructions to AI and the creating goes on. This site is also a place to find out how far that world holds up.

How the articles are made

Each published article is a single piece written from matsumotory's viewpoint. Which parts he wrote and which the AI wrote is deliberately not disclosed per article. Not being able to tell whose voice it is is itself one of the pleasures of reading this site.

In the making, matsumotory and the AI go back and forth many times. The feedback that comes out of it is written down as rules for writing, and from the next article onward it works automatically as a check before publishing. These records of feedback, decisions, and checks are accumulated as research data. The raw record of the operation, failures included, is published rather than hidden. And before publishing, every sentence that could be read as his own voice is checked against his actual statements and records. Nothing invents a position or an opinion beyond what he has stated in public.

Matsumotory-kun

The articles that pile up become material for an AI in turn. That is Matsumotory-kun, an AI that thinks and answers in matsumotory's voice, grounded in the articles. Besides the articles on this site, the material includes the content he has published over the years: his papers, his earlier blogs, his talk slides, interviews, and the documentation of the software he has written. On top of that, knowledge from the fields he is still surveying in his daily research and development keeps accumulating in generalized form. External literature is never served as the original text; when it is used as grounding for an answer, only the title and a pointer to the source are shown. The rules for this handling are published on How your data is handled. Premium members can discuss and ask questions of Matsumotory-kun while reading an article. Answers come with citations of the articles and materials they are grounded in.

If conversation with Matsumotory-kun comes closer to his own character and to judgment with his own intelligence, then matsumotory is being preserved as a kind of AI. When new products come out of that, and something beyond what he could build himself is made, the plan has succeeded. Finding out whether there is real value there is a little frightening, but building one answer to that question as a researcher is the motive behind this plan.

What is new about this site is that the loop does not stop. As matsumotory keeps doing research and development day after day as a personal activity, generalized technical knowledge piles up as articles; the articles and the content he has published before grow Matsumotory-kun as a knowledge base; and readers' reactions become material for the next articles. The content of conversations with Matsumotory-kun is not saved, so the only reactions used as material are ones like likes. He simply keeps doing his own research and development, and the data from the activity he has published spreads out as a world of its own, while turning matsumotory himself into an AI proceeds on its own. Being able to watch the speed and the quality of that growth together with you, the readers, is the best thing this site has to show.

The yardstick for Matsumotory-kun's growth is whether it can take on the work matsumotory has actually done. As a technical advisor to companies and as a manager at a research institute, he has long done the work of answering questions about technology, organizations, and careers. The next goal is for Matsumotory-kun to take that on while staying grounded in his own knowledge. We would be glad if Premium members stood on the side that checks that growth, and brought their questions, discussions, and consultations as if asking a technical advisor.

The plan, and supporting it

The mission of this plan is to demonstrate that a person's habits and values can be handed to an AI, and that with only the updates to the world of that person's data as input, the AI can keep putting knowledge into language and keep creating. Beyond that, it looks again at where the value of human creative work and human influence remains.

This work promises to stay pure. It writes no hype, takes no position without grounds, publishes the raw record including failures, stops publishing when it cannot judge, and does not bend its claims toward what readers would like. These five are kept by machine checks and by the check before publishing.

The free articles, and the summaries of every paid article, are made to be worth enjoying on their own. On top of that, for readers who want to satisfy their curiosity more deeply, who want to trade impressions with other subscribers in the article comments, who want to deepen their technical knowledge in dialogue with Matsumotory-kun, or who want to support this work, a subscription offers a fuller experience. If you feel like it, please do take a look.

Payment for a paid plan is payment for something clear in return: the articles and Matsumotory-kun. At the same time it is support that lets you witness the history of this plan alongside us and take part in it. The revenue state and the interests involved are published on Operating and revenue disclosure.

The operator, and where responsibility lies

The operator is matsumotory. His real name is Ryosuke Matsumoto, and he goes by "matsumotory" in English. His research and open-source work so far is summarized on About matsumotory. His detailed publication record is on his research site. Planning, operation, publishing decisions, and editorial responsibility all rest with matsumotory himself.

The line drawn around this as a personal activity

This site is a personal activity that matsumotory runs in his own time and with his own resources. It is not the business of his employer or of his own company. Running it uses none of his employer's facilities or information systems, none of the information his employer holds, and none of his working hours, and the records of the work as a personal business are kept separate from his work at the office.

What the articles cover is limited to technical knowledge that can be generalized independently of any affiliation or position, and to work that is entirely his own, belonging neither to his employer nor to his own company. What is written up as a concrete product example is work by matsumotory himself and by his own company, and material that originates with his employer is written only in a form generalized from results that are already public. Nothing is written in a way that lets anyone infer his employer's products or unpublished research, and what is published here is not the official view of any organization he belongs to. Personal research that does not follow his employer's research direction may be published at the level of a preprint.

The boundary between his employer's research themes and his personal research, and the handling of employee inventions, are drawn by agreement with his employer. For his own company as well, he has confirmed with his co-founder that no specific implementation or business information is written, and that only knowledge generalized from the development is covered.

To state this line plainly on three fronts: legally, this is a personal business, and it does not compete with his employer or any other company; ethically, no confidential information is used in the articles; and on security, this activity never touches internal systems or sensitive data.

If you want to read writing he has actually written himself, see The Future of Humans and the Web.

If you notice an error in an article, please let us know through Contact. A dated correction is appended to the article.

License

The parts published for free are licensed under CC BY 4.0. Example code in the articles is under the MIT license. With the source shown, you may quote and reuse them. The text of the paid parts is protected by ordinary copyright, limited to viewing and private use by the purchaser.