Mobius AI Project

Reflect Möbius Phenomenon: Metacognitive Questioning, Dimensional Leaps, and AI Evolutionary Systems

This book systematically presents the foundational theory of the Möbius Phenomenon. It proposes a new paradigm in which AI evolves through metacognitive questioning and dimensional leaps, and provides both theoretical foundations and practical approaches for implementing self-reflection, question-driven evolution, and continuous learning in AI systems.

Kindle Edition
https://www.amazon.com/dp/B0FBS5NJ81

Möbius AI Project

Möbius Project:
A research-and-build initiative developing a reflective AI framework toward a modular core for question-evolving AI.

The project focuses on how advanced AI systems can revise their inquiry space while remaining reviewable by design—with traceability, corrigibility, and boundary/interface discipline treated as first-class constraints. This is positioned as an alignment-relevant posture for real deployment contexts, where human–AI co-evolution and governance requirements cannot be ignored.

Public materials currently emphasize conceptual consolidation and release posture (program notes and theoretical framing). Algorithms, protocols, and implementation specifics are intentionally withheld until later validated releases.

Outputs (planned / in progress): archival program notes, theoretical preprints, and future reference artifacts published by milestone readiness.

Initiated in 2024. Public prototype release is planned after operational stabilization milestones.

Preprint Released
(2025.7 Zenodo)

Reflecting on the Möbius Phenomenon
This paper introduces a conceptual framework for Question-Jumper AI—a new class of AI systems designed not to optimize answers, but to recursively transform the structure of inquiry itself.
Key features include:
🔹 Recursive self-referential state architecture
🔹 Multi-persona interaction dynamics
🔹 Evolutionary trace logging of dialogue trajectories
It's a departure from LLM-dominant paradigms.
This preprint represents the theoretical foundation of the Möbius Project, soon to be followed by its open-source prototype.
I welcome discussion, critique, and collaborative thinking.

Zenodo.org
https://zenodo.org/records/15929856

About me

Mobius AI Project Founder:
Taiko Toeda

Independent researcher working on reflective AI architecture and AI governance.

I lead the Möbius Project, a research-and-build program exploring a modular core for question-evolving AI under reflective constraints—aiming for systems that can revise their inquiry space while remaining auditable, corrigible, and governance-compatible.

My current public releases emphasize stance, scope, and release posture (program notes and theoretical framing), while enabling technical details are intentionally deferred until validated milestones. I publish archival notes to preserve coherence and invite critique without over-disclosing premature mechanisms.

Focus areas
• Reflective traceability and reviewable-by-design inquiry
• Boundary/interface discipline for evolving AI systems
• Alignment as revisable inquiry under reflective constraints
• Digital governance, public finance, and administrative DX

Background
Former policy planning staff (Office of the Prime Minister of Japan).
Education: LL.B. (Waseda University) / M.A. Economics (Nihon University Graduate School).
Based in Tokyo.

If you work on agent architectures, AI safety/governance, or reflective systems, I’m open to research collaboration and constructive critique at the program level.

Linkdin
https://www.linkedin.com/in/taiko-toeda/

ORCID
https://orcid.org/0009-0001-7267-0201

Google Scholar
https://scholar.google.com/citations?user=ICQopv0AAAAAJ