Arkadium
Globalística notebooks, manifesto, academic papers and pedagogical manuals on the Meta-Globàlium and artificial wisdom.
This page gathers the published documents of the Arkadium project. It spans the pedagogical line opened by Lluís M. Xirinacs with the Globàlium (1997), its current continuation as the Meta-Globàlium, and the founding documents of the present programme: the position manifesto, the academic paper and the in-depth notebooks.
All texts are published under free licences (CC BY-SA 4.0 or CC BY-NC-ND 4.0, as indicated in each document) and may be quoted, translated and redistributed within the terms of the respective licence. Most documents are written in Catalan; English versions are indicated where available.
The founding position of the project. It argues that a more humane AI is possible, structurally anchored to a global model of reality. Six theses: structural verifier, top-down interpretability, neuro-symbolic architecture, sovereignty, functional wisdom, and a cooperative-tool stance towards AGI.
The academic formalisation of the Arkadium architecture: a neuro-symbolic agent that anchors its inferences to the Meta-Globàlium for the structural verification of human judgment in AI systems. Includes the mathematical formalisation, the validation method and a worked use case.
The six volta methods — Application, Revelation, Culture, Universe, Relation and Consistency — with their eight-station circuits, the four scales of application, and a step-by-step implementation guide for organisations and communities that want to develop their potential. Includes the application of the methods to Arkadium and to Kosmos.coop. In Catalan.
Documents for developers and integrators: the agent architecture, the end-to-end pipeline of a response, the design of the wisdom metric 𝓦, the scope parameter and the structural chips of the interface.
A structural evolution of the global model of reality. It walks through the most significant differences between Xirinacs' Globàlium (1997) and the Meta-Globàlium (2025–2026): principles on the OBJ–SUB axis (identity, non-contradiction, alterity), tempeternity, the fractal N1/N2 architecture, axiomatic operations, the Aristotelian mesótes and the solve-coagula dynamic. In Catalan.
Notes on the possibility of a wise AI, aligned with human values and anchored in a global model of reality. It explores the opportunities and risks of AI, the transition towards an artificial wisdom via the Globàlium, and the auditable inference cycle. Includes a 2026 update note on what has changed since the original writing. In Catalan.
The official manual of the Globalística workshops at Mas el Negre. A didactic presentation of the Globàlium: the eight areas, the twenty-six second-level categories, the eighty third-level ones, the six structural characteristics and the three voltes of the Application Method. Includes the four globalist skills that the practice develops. In Catalan.
The reference glossary of the structural verifier: the 𝓗 (harmonic completeness) and 𝓦 (wisdom score) metrics with their seven components, the eight cardinal poles of the Meta-Globàlium, the twelve mediators, the three cycles of the Application Method and the diagnostic metric ℑ (implicit coverage). Includes a pedagogical layer with quick readings and a mapping to the ANA→SIN→AMO→EXP phases.
An empirical comparison of the structural verifier over three test domains. It discriminates adversarial reward-hacking answers (Lorem ipsum + decorative cardinal codes) from genuinely dialectical answers by a factor of ≥8× on the wisdom metric 𝓦. Results reproducible from the public verifier code (Apache 2.0).
The founding doctoral thesis of the Globàlium — the crystallisation of more than forty years of Xirinacs' philosophical and topological research. It is the reference text of the original model on which the Meta-Globàlium has been built.
