Takuya Nemoto

About me

I am enrolled at ZEN University, Faculty of Social Informatics — in its first autumn-entry cohort. The programme spans the humanities and the sciences.

My coursework is mainly philosophy and business practice. Mathematics is on offer as well. The courses I have credit for are listed in the CV (in Japanese). What this place holds is my study of computer science and artificial intelligence.

What is here rests on self-study and on working with language models. There was no supervisor and no collaborator. “Independent Researcher” is a provisional title used for clarity; in fact the author is an undergraduate. The university credits, though, are not self-study. They came from taking courses, submitting work and being assessed. They are what the table below lists, and what self-study and language models refer to lies outside them — the three papers in philosophy, the three Trinity-Infinity papers, and the tools and records kept here.

This site itself is built with Claude, by Anthropic. The page generation, the checking tools, the records — the git history carries Co-Authored-By: Claude Opus 5 throughout. Which tools went into the three philosophy papers, though, cannot be identified — no record was kept and the author’s memory does not settle it (E5 in the errata for the three). On the Trinity-Infinity side it can be identified — the 2025 first draft was generated by a language model (ChatGPT, OpenAI), and the 2026 revision was written with Claude Code (Anthropic). The three philosophy papers are the only ones where the tool cannot be pinned to a single name.

The bar that was cleared, in the venues' own wording. SSRN — the subject must be part of the world-wide scholarly discourse covered by one or more of SSRN's subject area networks; it must not be non-scholarly, articles with no references being named; title, date written, abstract and full author details must be present; it must clear research-integrity checks. Editorial staff look at it before it goes up. PhilArchive — it takes works of all types (articles, books, dissertations) in all areas of academic philosophy; non-philosophy only where cross-disciplinary and of clear interest to philosophers; All books and papers submitted should be of professional quality; threats and harassment, personal information without consent and impersonation are prohibited; the right to reject any submissions is reserved; and more than six hundred category editors keep irrelevant or misfiled items out of their categories — which acts after the fact, not before.

What follows from that is definite. Work written by self-study and language models was judged by SSRN's editorial staff to belong to the scholarly discourse of its field. For two papers, a person looked before publication and did not turn them away. The third was turned away. Fragmentarian Spiritual Individualism (10.5281/zenodo.22064241) was submitted to SSRN and rejected. This gate does turn work away. The field is named, not left vague: the two philosophy papers that were accepted sit in Continental Philosophy in the Philosophy Research Network, and Trinity-Infinity Series II in Numerical Analysis in the Mathematics Research Network. The source note on the naval gazette entry was also accepted by SSRN's editorial staff, and sits in East Asian History in the History Research Network (SSRN 10.2139/ssrn.7449338). The philosophy-of-science paper Machine-mediated Rediscovery was also accepted by SSRN’s editorial staff (SSRN 10.2139/ssrn.7537983). On SSRN a subject area is itself an eJournal. The rejected paper was submitted to the same Continental Philosophy area. It was not accepted, so it sits in no eJournal at all. On PhilArchive, two of the three were taken to lie within academic philosophy and to be of professional quality. The right of rejection is reserved; for those two it was not exercised. The same thing has happened on the mathematics side: Trinity-Infinity Series II was judged by SSRN’s editorial staff to belong to the scholarly discourse of its field (SSRN 10.2139/ssrn.7446961). For one paper, a person looked before publication and did not turn it away. Series I (revised) and Series III were submitted and turned away. Two notices arrived from SSRN on 2026-09-18. The wording is boilerplate: the submission “does not meet SSRN’s submission requirements or falls outside the scope of our platform”, and it does not say which. One thing is certain here — Series II sits in the same Mathematics Research Network, so mathematics being out of scope cannot be the explanation. What separates the two that were turned away from the one that was taken is not yet known. So this gate sorts within a single series. Of the eight submitted, five were accepted and three were turned away.

What does not follow is equally definite. Neither gate looked at whether the argument holds. What was passed is a gate on acceptance, not peer review; on the PhilArchive side two of the three passed a gate and one passed none at all. The same holds for Series II: what SSRN looked at was subject and form. The rejection is the same coin, the other face. All that follows from it is that SSRN did not accept those papers — not that the argument was found wanting. The platform says so itself — the notice sent for the two mathematics papers states that this decision does not reflect a judgement on the merits of your work. For Fragmentarian no reason was given. For the two mathematics papers a boilerplate notice arrived, but it does not say whether the ground was requirements or scope. Neither has been established here. All of the wording is set out verbatim above.

One thing about the mathematics is now settled. The operator behind the three Trinity-Infinity papers, x ← DQx + (I−D)p, was already a known model — the Friedkin–Johnsen model (1990 / 1999) in sociology. The opinion-dynamics literature routinely names DeGroot and Friedkin–Johnsen together as the most influential models in that area; it became a Cambridge monograph in 2011, reached Science in 2016, and is now used in control theory and computer science. Not a footnote — one of the two pillars of the field. And Series I names consensus dynamics in its own first section while citing nothing from that field (ERRATA E8, severity high). The dimensions, stated exactly: this operator is the special case where the influence matrix W is a cyclic permutation — a ring in which each party listens to exactly one neighbour. The general W, who influences whom and by how much, which is what makes Friedkin–Johnsen sociology, is dropped entirely. What was rediscovered is the mathematical frame of a first-rank sociological model, with the sociology taken out, and its load-bearing term.

One thing has been added beyond that. On 2026-10-04 a note was written showing that the same operator decides the equilibria of a repeated game (trinity-operator, games). The continuation values of a cyclic path of play are the fixed point of this operator. The formula itself is the textbook one. From it, thresholds were proved: a path on which three players take turns defecting is an equilibrium exactly when the discount factor is at least 1/φ. The note also shows that the Series II example (2,4,2) is not sustainable for any discount factor. The propositions have not been checked against the literature. They may be known exercises, and they are not counted as new results until they are checked. The note also states a conjecture: without randomization, the set of sustainable payoffs is full of holes. It is called the golden-window conjecture, the window being the discount factors from 1/φ to 2/3. The name describes the statement and is not a person’s name. The conjecture and what is proved outside the window are set out on a page of their own. There is no proof, and the evidence is numerical only. The known model, the proved propositions and the conjectures are written down separately.

The preprints up to this point were written as a hobby. None of them has been peer-reviewed. This is not modesty. It follows from what is set out above — the desk gate was passed, but no one has examined the arguments. The three mathematical papers are a special case of a known model, and the contribution to the field is zero. Hobby research ended in the first undergraduate year. What comes next will be written so that the word need not apply.

The edX and Tohoku University MOOC certificates were also finished in the first half of the first undergraduate year. There are seven of them. And everything set out here is the work of that first undergraduate year. Nine preprints, 16 Zenodo DOIs, ten repositories, seven certificates and open badges. The dates themselves have been checked against nothing. They are printed on the certificates, but neither edX nor the issuers can be reached from this working environment. The enrolment date is not printed here either. Printing it would mean printing a date that has not been verified.

Where it goes is decided too. Anything taken this far by self-study goes only to diamond open-access venues — venues that charge neither the reader nor the writer. No publication fees, no submission fees. Solicitations that advertise being free are declined as well — the moment money changes hands, what passed was payment, not selection. The rule has already been used once: a publisher wrote offering to turn the work into a book, and no reply was sent (solicitations received). That was not a judgment about them — their existence, record and review practice are all unverified here. What was decided was a rule on this side.

How published work is counted is fixed as well. Only a degree thesis, or work in a journal ranked Q1 within its field category, or in a peer-reviewed diamond open access journal in the field, goes into the article sections of researchmap and HAL. Everything else is treated as a working paper, a preprint, or MISC. The floor for that is appearing on SSRN, PhilPapers, or arXiv. Work that exists only on Zenodo goes into neither section. Zenodo is treated as the place where work that did not pass those screenings is kept for the sake of research transparency. The Zenodo DOI remains the canonical one to cite. What is on SSRN stays outside the article section too. Nothing qualifies at present. The rule above on where work is sent does not change; this one only decides how what appears is counted.

Apart from those two, there is a recurring goal. Acceptance in the Journal of Open Source Software (JOSS, ISSN 2475-9066). One acceptance will not close it. For some time to come it is set as a goal to be repeated. It charges neither publication nor submission fees — it falls on the side rule 14 allows. On acceptance a Crossref DOI is minted. The prefix is 10.21105, giving DOIs of the form 10.21105/joss.03021. The stated requirements, in the venue’s own wording. The software must be open source as per the OSI definition. there should be an actual license file present in the repository not just a reference to the license. The software must be hosted at a location where users can browse the source code files, open issues, and propose code changes without manual approval of (or payment for) accounts. The software must have an obvious research application. You must be a major contributor to the software you are submitting. feature-complete (i.e., no half-baked solutions). Four gates come before review. Failing any one returns the submission unreviewed (will receive a desk rejection). One, the length of public history. The repository must have been public for more than six months prior to submission, with active development spanning that period. Two, evidence of research use. There must be evidence that the software is being used for research, and Aspirational statements about future use are not sufficient — saying it will be used is not enough. Three, open-source practice. For a single author, several indicators must be present at once: a public commit history over time, tagged releases or a changelog, tests and CI, documentation, a CONTRIBUTING file, and stated support expectations. Four, iterative development. The development history must show ongoing iteration, not a single burst of commits. The dates are fixed. The earliest first commit across the ten repositories is 2026-06-12, which is this site itself; on the tooling side the earliest is 2026-09-05. Counting back from the six-month requirement, the earliest possible submission is 2026-12-13, and 2027-03-06 on the tooling side. For errata-check, the one to be submitted, the first commit is 2026-09-07 and the earliest possible submission is 2027-03-08. The concentration of work into a few weeks is not filled in by waiting; it needs a record of continued work while the tool is in use. Submission comes after that. The venue also requires disclosure of language-model use (AI usage disclosure) — the tools and versions, where they were used, and confirmation that a human reviewed the output. Rule 8 decided the same thing first. Only errata-check is to be submitted. The others are papers, a source note, a memorandum, or tools judged not to reach the scope of JOSS. Which gate each one falls short of is set out one by one in the ten repositories. The wording is copied from docs/ in openjournals/joss, with only link and emphasis markup removed.

The papers are being submitted too. Both the three philosophy papers and the three Trinity-Infinity papers are going to diamond open-access refereed journals. The decision changed on 2026-09-15. Until then they were not to be submitted, and the reason was not cost: the use of language models could not be disclosed in full. Journals ask for the tools and versions, where they were used, and confirmation that a human verified the output. For the three philosophy papers the translation tool cannot be identified — no record was kept and the author’s memory does not settle it (E5 in the errata). The reason for the change is one point. Writing that the tool cannot be reconstructed is writing everything that is known. What is incomplete is not the disclosure but the material available to disclose. Whether that is acceptable is the venue’s decision, not one to pre-empt by self-rejection. The COPE rules are kept — no language model is listed as an author, and the use of language models is disclosed. The preprints do not move. They remain on Zenodo, PhilArchive and SSRN. This is not a withdrawal, and it is separate from submitting to a journal. One box will not be ticked: any attestation that the writing is substantially the author’s own (the ten repositories).

The direction has been changed. Decided on 2026-09-23. From here on the effort goes into producing results inside the institution. Presenting at conferences, and being awarded there. Neither has happened even once. What exists so far is preprints that passed a desk gate, and deposit records. Neither of those is a peer-reviewed result. Setting JOSS as a recurring goal points the same way.

The outlook, stated plainly. Nothing outside those two is in view — working as an independent researcher after a doctorate, and earning a living as a sole proprietor who takes on only work chosen for its quality. Conferences and awards lie on the way to the first of those. The count of goals has not become three. A doctorate, too, is in essence non-reproductive: it is conferred on one person’s work alone and can be neither handed over nor inherited. And life will be lived with contact with others cut back as far as it will go. That way of living points the same way as the writing. What The Nobility and Exemplarity of the Celibate Individual defends is precisely that form — a happiness dependent neither on others nor on the cycle of desire (C2 in that paper). Whether that form can actually be taken up, however, is not established. C2 rests on two sources the author has not read (E6 in the errata, severity high), and the statistics on isolation press exactly on whether it can be taken up. A way of living is not evidence for a claim. The order runs the other way: the claim is the part not yet standing on its own feet.

What is to remain after death (the gravestone and posthumous name, the degree, ORCID and DOI) is not research, and has been moved to a separate page.

1. Tochigi Prefectural Tochigi High School (120th cohort)left partway through
2. Sundai Preparatory School, Omiya, course for high-school graduatesattended
3. ZEN University, Faculty of Social Informatics (1st cohort)enrolled (autumn entry)17 courses, 34 credits — 14 required, 16 required elective, 4 elective (as of 17 September 2026)
   • Introduction to the Humanities and Social SciencesrequiredB
   • IT LiteracyrequiredC
   • Academic LiteracyrequiredP
   • Using Digital ToolsrequiredB
   • Multilingual IT CommunicationrequiredB
   • Applied Practice in Artificial IntelligencerequiredC
   • Introduction to EconomicsrequiredC
   • Introduction to Information Securityrequired electiveC
   • History of Mathematicsrequired electiveB
   • Introduction to Publishing on the InternetelectiveA
   • Internet Culture through the History of Derivative WorkselectiveB
   • A History of the Japanese Game Industryrequired electiveC
   • Business Managementrequired electiveC
   • Law Irequired electiveC
   • Psychologyrequired electiveD
   • Applied Machine Translation (Informatics)required electiveD
   • Applied Machine Translation (Law)required electiveD

The English course titles below are unofficial translations, not the university's own. The official English titles have not been checked — the syllabus cannot be opened from the environment this site is built in. The faculty name is the form the university's English pages are reported to use; those pages present a single faculty with no department level, and they could not be opened from here either.

Emblems of standing and turning alone, and of non-reproduction

• To work as an independent researcher after a doctorate
• To earn a living as a sole proprietor, taking on only work chosen for its quality