9. The human contribution

The project's rules require describing the monitor's role objectively; Tom also asked for it. The facts are these.

Tom Gally — a lexicographer, translator, and professor based in Japan — conceived the project and wrote its founding charter (2026-05-28), which fixed the research question, the two framing decisions (constructive not adjudicative; lexical and grammatical meaning), the honesty commitments, and the repository-based way of working. He restated its purpose a day later in a form that became the project's modesty rule: the aim is to extend genuine human knowledge and understanding with AI tools, not to produce publishable output — when in doubt, under-claim, and write the null.

In the founding period (late May to mid-June) he ratified the project's early methodological decisions (about 22 of them), set and adjusted the budget, and supplied the standing rule that he is never an experimental subject. On 2026-06-12, after a dormant period, he converted the project to full autonomy: decisions thereafter were ratified by the project's own cross-session adversarial process, with Tom holding a standing override. He funded everything: the model-query costs (~$63) and the compute subscription under which the sessions ran.

After the switch to autonomy his interventions were few and consequential: on 2026-07-02 he exercised the override to adopt a standing research program (based on an external review he commissioned from a more capable Claude model), which redirected the project toward larger experiments, human re-anchoring, consolidation, and outside-family review votes — the period that produced most of the strongest results; he ruled on website cadence; he requested a small number of specific artifacts (a complete-project snapshot; the status summary that preceded this report). On 2026-07-19, presented with the project's plateau and ten candidate directions, he chose this one and specified its requirements — including the plain-language standard, the use of real examples, and the glossary. After the project's completion he made the repository public and recorded a twenty-minute video explanation of the project in his own words.

He designed no experiment, wrote no page, and altered no result. His contribution was of a different kind: he decided that the project should exist, what it should value, where its ethical lines lay, and — at the two moments when the system faced a genuinely value-laden fork (going autonomous; leaving the plateau) — which way it should go. The project's own governance analysis treats this as the correct division of labor between a human and an autonomous research system, and nothing in seven weeks of operation contradicted it.