Douglas G. Stevenson, Civilian Scientist, United States of America
tevenson-Flux Information Theory (SFIT) proposes that gravity is a dynamic, informationcarrying flux oscillating at the geometric resonance frequency νres = 1.20134 mHz with coupling kernel K = 1.060. This non-reciprocal metric perturbation couples classical gravity to quantum systems and produces observable laboratory signatures, including a 14.28σ modulation in ultra-cold neutron data, Kohlrausch–Williams–Watts (KWW) relaxation tails with β = K, and quantized information flux analogous to M2-brane charge quantization in M-theory. We present the complete mathematical framework, derive the flux quantization condition from single-valuedness of the neutron wave function, establish the information-theoretic origin of the KWW stretched exponential, and show how SFIT provides a low-energy effective realization of holographic and ER=EPR principles. The theory is fully laboratory-testable and offers a new information-theoretic bridge between general relativity and quantum mechanics. All data, code, and synthetic event generators are openly available.
Information Theory, Resonant Flux, Non-Reciprocal Metric, KWW Relax-ation, Flux Quantization, Holographic Duality, Quantum Gravity, Ultra-Cold Neutrons, Machine Learning Applications
Jered McClain , SEXA Institute of Technology and Interdomain Galactic Advisors,United States of America
This paper formulates Recursive Information Admissibility (RIA), a traceable constraint architecture for multivariate information validation, fault localization, and loss-aware dimensional reduction. Prior mechanical evidence is extended with a controlled four-task information benchmark and an independent public-data novelty study. In the controlled suite, RIA reduced macro false admission to 35.08%, outperforming LOF at 42.75%. Across 17 public novelty tasks and 306 evaluations, however, generic RIA did not lead: false admission was 78.61% versus 73.61% for Isolation Forest, while LOF achieved higher ROC-AUC. The combined result narrows the supported claim: RIA is promising when domain relationships and invariants are explicitly known, but is not demonstrated as a universally superior anomaly detector.
Recursive Information Admissibility, Information Validation, Artificial Intelligence, Knowledge, Representation, Information Integrity
Arman Shahpari and Arezou Sadoughi, Appalachian State University, USA
Artificial intelligence (AI) is increasingly discussed as a participant in the construction workforce rather than merely a software tool. This paper conducts a systematic critical review of AI applications across the construction lifecycle — pre-construction, construction, and operations/post-construction — and proposes an “Augmented Workforce” taxonomy that classifies AI systems by their degree of autonomy relative to human tasks (Human-only → AI-assisted → AI-augmented → Human-supervised AI autonomy → High-autonomy task execution). Using a PRISMA-style review protocol, we screened records from Scopus, Web of Science, Engineering Village, ScienceDirect, and Google Scholar (search window and terms reported in Section 2), and we introduce a five-level evidence hierarchy (A–E) to separate independently validated academic findings from vendor-reported claims. We further propose a quantitative validation framework linking AI performance metrics (precision, recall, F1-score, mAP) to construction-management KPIs (schedule variance, cost variance, rework rate, detection lead time). Findings indicate that current evidence robustly supports task-level augmentation but does not yet establish broad autonomous workforce substitution. We identify gaps in reproducible benchmarking, economic analysis, and governance research, and propose a structured research agenda.
artificial intelligence, construction management, augmented workforce, systematic review, evidence hierarchy, quantitative validation framework
David Allan Nelms, Independent Researcher, Inverness, Mississippi, USA
Extended language-model workflows can lose task constraints, reuse superseded decisions, promote hy-potheses into facts, or misreport completed work. This paper specifies the ATLAS Autonomous Drift Con-tainment System, an architecture that treats these failures as unauthorized state transitions. Its components are a versioned project baseline, typed evidence records, authority-aware updates, independent checks, controlled repair, and recoverable checkpoints. The contribution is a reproducible transition protocol and evaluation design, rather than a new memory mechanism or a claim of universal reliability. A conditional invariant-preservation argument defines the controller’s enforcement boundary. A small deterministic im-plementation exercises ten scripted cases and additional recovery checks; it demonstrates conformance of those rules without evaluating an LLM. A controlled study is proposed to measure drift, legitimate-update acceptance, task completion, and cost against matched baselines. Effectiveness in language-model work-flows remains unvalidated.
Large language models, Operational drift, State management, Provenance, AI reliability
Copyright © ITEORY 2026
