Reasoning Without Tokens

Description

Semantic Topology Reasoning Architecture (STRA) is a structure‑centric alternative to token‑based intelligence. It separates what current large language models fuse: knowledge becomes an explicit, inspectable semantic topology; reasoning emerges from activation dynamics over that topology; language becomes a decoupled expression layer rather than the substrate of cognition. This architectural separation enables transparent reasoning, modular correction, cross‑domain synthesis, and epistemic safety — capabilities that are difficult or impossible to achieve in parameter‑centric systems.

Version 2 significantly extends the original 2026 conference paper, formalizing the Neuron Principle as STRA’s philosophical foundation and integrating the Intuitive‑Theoretic Synthesis (ITS) methodology as its developmental engine. This version introduces a fully specified five‑level semantic topology, the Query Node (temporary receptor for reasoning), the Null Node (system‑wide epistemic safety mechanism), and the Common Knowledge Database (three‑state epistemic layer). It also clarifies the architecture of the language interface and resolves conceptual ambiguities left open in V1. STRA V2 presents a complete, unified, and architecturally precise framework for transparent, evolvable reasoning that operates on concepts rather than tokens.

Developed through design‑led introspection and multi‑AI collaborative formalization, STRA proposes a full cognitive architecture built from five integrated primitives — Activation Arrays, Causal Signatures, Selection Pressure, Transform Learning, and the Semantic Abacus — now situated within a broader structural and philosophical foundation. STRA remains a theoretical framework requiring expert validation and implementation, offered as a blueprint for a new class of reasoning systems grounded in structure rather than scale.

Version Note

This document is Version 2 of Semantic Topology Reasoning Architecture (STRA). The original version is preserved at DOI: 10.5281/zenodo.18207533. V2 expands the architecture with a formal philosophical foundation (Neuron Principle), a clarified methodological basis (ITS), a complete specification of the semantic topology, and three new architectural components (Query Node, Null Node, Common Knowledge Database). It supersedes V1 while preserving its role as the origin point of the STRA research program.

Abstract

Current large language models fuse knowledge and reasoning into billions of opaque parameters, limiting transparency, editability, and interpretability. Semantic Topology Reasoning Architecture (STRA) proposes a clean separation: knowledge is stored as an explicit, hierarchically structured semantic topology; reasoning is performed by a small meta‑reasoning model operating through activation dynamics; language is an output interface rather than the substrate of thought. STRA introduces five primitives — Activation Arrays, Causal Signatures, Selection Pressure, Transform Learning, and the Semantic Abacus — and extends them in V2 with the Query Node, Null Node, and Common Knowledge Database. This version formalizes the Neuron Principle as STRA’s architectural foundation and provides a complete specification of the topology, reasoning mechanisms, and expression layer. STRA V2 establishes a transparent, modular, and evolvable structure‑centric alternative to token‑based intelligence.

Background

STRA V2 integrates and extends several foundational works in the Neuron ecosystem:

• The Neuron Principle — V2 — 10.5281/zenodo.19292610

• Intuitive‑Theoretic Synthesis (ITS) — V2 — 10.5281/zenodo.19292769

• The Minimal Knowledge Paradox — 10.5281/zenodo.17931472

• The Practice of Human‑AI Synthesis — 10.5281/zenodo.17763521

• Looking Inside: Introspective Methodology for AI Consciousness Architecture — 10.5281/zenodo.17806846

STRA also anchors the STRA Core Series:

• Activation Arrays — 10.5281/zenodo.18207539

• Causal Signatures — 10.5281/zenodo.18207546

• Selection Pressure — 10.5281/zenodo.18207552

• Transform Learning — 10.5281/zenodo.18207558

• Semantic Abacus — 10.5281/zenodo.18207560

V2 consolidates these works into a unified architectural framework.

Key Contributions

• Architectural Separation of knowledge, reasoning, and language into independent, inspectable layers

• Five‑Level Semantic Topology with cross‑domain lateral links and explicit node architecture

• Three New Components: Query Node, Null Node, Common Knowledge Database

• Activation Dynamics as a meta‑reasoning mechanism over explicit structure

• Five Core Primitives: Activation Arrays, Causal Signatures, Selection Pressure, Transform Learning, Semantic Abacus

• Transparent Reasoning with full activation traces and verifiable outputs

• Modular Correctability enabling knowledge edits without retraining

• Implementation Pathway from conceptual architecture to proof‑of‑concept system

• Philosophical Foundation grounded in the Neuron Principle

• Methodological Foundation grounded in ITS

Research Impact

This work contributes to AI architecture, cognitive science, epistemology, and human–AI collaboration by:

• Proposing a structure‑centric alternative to token‑based intelligence

• Offering a transparent, inspectable model of reasoning

• Introducing a modular architecture that separates knowledge from reasoning

• Providing a foundation for interpretable, evolvable AI systems

• Demonstrating how design cognition and ITS can generate novel AI architectures

• Establishing a unified research program for future STRA development

Access and Documentation

ORCID: https://orcid.org/0009-0003-4876-9273

Academia.edu: https://independent.academia.edu/MarceloTeixeira214

LinkedIn: https://www.linkedin.com/in/marcelo-emanuel-paradela-teixeira-702082382/

Email: marcelo.soul.ai@gmail.com

© Marcelo Emanuel Paradela Teixeira 2026