# Human–AI note This research is done by a human and AI working together. This note separates who contributed what. ## What I did - **Ilanguage, the description language.** The idea of describing physics with one object-oriented description language (objects, states, getters and setters, ownership, contracts) and the language itself are my work. Its foundations were laid in an earlier project, [oop-physics-language](https://github.com/ivanhonis/oop-physics-language). It has been refined since; the research published here uses version 1.b. The language went through several unsuccessful versions. Which of them were dead ends, which statements matter and what stays in the language were my decisions. The idea draws on known research directions; the decision that such a language should exist, and that it frames the whole work, is mine. - **The engine of the language.** The earlier versions described states but did not say what drives the system. The AI did not see this from the partial successes and failures. I identified that a driving engine was missing, and I defined the solution: the contract not only scores the states, it drives the evolution. This is inspired by existing physics; the diagnosis and the decision are mine. - **The research directions.** Which questions are asked, and along which route, I decide. For example: that space is to be derived through the view of a part (subproject 03), and, for the gravity programme to come, that gravity is not a single component but an emergent interplay of modules. The question is then not which module is gravity, but whether the modules together form a closed, stable, mutually constraining system. - **Leading the research.** Setting up the subprojects and their success criteria, approving or rejecting packages, accepting results. - **The methodology.** The package-based workflow: one question per package, versioned derivations, a dependency graph with statuses, limited regenerations, numerical checks fixed in advance, cross-verification by AI models of different providers, and closing reports. I designed it and I enforce it. Its principle: no AI checks its own work, and no AI's claim is accepted on the word of another. ## What the AI did - **Working out the packages.** The derivations, the detailed arguments and calculations, most of the mathematics and of the wording, under my direction and with my approval. - **Physics background.** Bringing in the knowledge of the literature, and relating the results to existing physics. - **Objections and checks.** Attacking its own results: consistency checks, counterexamples, negative results. - **Cross-verification.** The derivations are checked by AI models of different providers: a derivation is reviewed by a model from another provider than the one that wrote it, which sees only the question, the derivation and its code. The reviews, their verdicts and the corrections they led to are published with each package. ## Limitations I am an AI engineer. Physics is not my primary field; my mathematics is that of an AI engineer. I cannot check every derivation in full depth on my own. The cross-verification by models of different providers mitigates this, but it does not replace review by human experts. --- This is not a closed, proven theory but a structured research programme that keeps its cards open: every result carries its status, earlier versions and external reviews stay visible, and criticism from experts is welcome. Ivan Honis