The corpus

Where the work lives.

The DFT corpus has three pillars. A 25-chapter book that develops the theory from first principles. Sixteen papers that treat focused topics in depth. The dft-solver, a Python implementation of the numerical machinery. A 28-claim patent draft sits alongside, covering the algorithmic and hardware contributions. Nothing here claims finished status; every piece is open to external scrutiny.

01

PAPERS

Sixteen manuscripts.

The geometric-framework paper on Riemannian-manifold learning has been accepted in Frontiers in Big Data. The remaining manuscripts are in preparation; pre-publication drafts are available on request for academic peer review. Each row below names the question the paper takes up; the full title and the publication target sit below.

  • P-01 The Parrondo Paradox as a Non-Equilibrium Phase Transition in Data Field Theory non-equilibriumalgorithms In prep
  • P-02 Data Field Theory: An Axiomatic Framework for Intelligence with Testable Physical Signatures foundationsaxiomatization In prep
  • P-03 Data Field Theory: A Geometric Framework for Learning on Riemannian Manifolds with Synthetic Validation and Limitation Analysis Frontiers in Big Data · DOI 10.3389/fdata.2026.1752468 foundationsempirics Published
  • P-04 The Mass-Robustness Law: A Field-Theoretic Principle for Generalization, Mathematical Foundations and Experimental Validation robustnesstheory + empirics In prep
  • P-05 Advanced Mathematical Extensions of Data Field Theory: Hyperbolic Geometry, Gauge Theory, and Higher Structures geometryfoundations In prep
  • P-06 Energy-Based Learning as Field Dynamics: Geometric Foundations and Symmetry-Aware Filtering in Data Field Theory energyalgorithms In prep
  • P-07 The Data Field Processor: A Computing Stack for Field-Theoretic Learning hardware In prep
  • P-08 Universal Critical Phenomena in Learning Systems: Comprehensive Experimental Validation of Data Field Theory Across Domains universalityempirics In prep
  • P-09 Structural Analogies Between Data Field Theory and Quantum Field Theory: A Mathematical Framework for Cross-Domain Method Transfer philosophydiscipline In prep
  • P-10 Hierarchical Parrondo Paradoxes in Data Field Theory: Rigorous Framework, Validated Scaling, and Design Principles algorithmsmulti-scale In prep
  • P-11 Data Field Thermodynamics: A Unified Framework for Intelligence, through Geometric Field Theory and Thermodynamic Principles thermodynamicsfoundations In prep
  • P-12 Data Field Theory: A Unified Mathematical Framework for Physics-Grade Intelligence synthesis In prep
  • P-13 Unified Theory of Critical Learning and Data Field Processing: Bridging Statistical Physics with Specialized Computing Architecture theory + hardware In prep
  • P-14 Advanced Mathematical Foundations of the Mass-Robustness Law: Hyperbolic Geometry, Gauge Theory, and Spectral Analysis foundationsrigour In prep
  • P-15 Quantum-Enhanced Hierarchical Systems: A Field-Theoretic Framework for Emergent Optimization quantumalgorithms In prep
  • P-16 Data Field Theory: A Geometric and Field-Theoretic Foundation for Emergent Intelligence foundationsbaseline In prep

02

THE BOOK

Twenty-five chapters.

From the diagnosis of modern AI's heuristic plateau to the empirical-validation chapter, the book moves through the mathematical foundations, the theory, the numerical methods, the empirics, the applications, the safety layer, and the integration. Six AGI-variant chapters (CH19AGI–CH24AGI) repackage chapters 19–24 for a different audience.

ChapterTitlePart
CH1Introduction: The Quest for a Physics of IntelligenceI. Diagnosis
CH2Mathematical Foundations: Manifolds, Fields, and OperatorsII. Mathematical foundations
CH3Physics Primer: Field Theory and Critical PhenomenaII. Mathematical foundations
CH4Core Axioms of Data Field TheoryIII. The theory
CH5The Data Field and its Manifold SubstrateIII. The theory
CH6Dynamical Laws: From Ginzburg-Landau to Generic Learning FlowsIII. The theory
CH7Observables in DFT: How to Measure "Intelligent" BehaviorIII. The theory
CH8Symmetry and Equivariance in Field DynamicsIII. The theory
CH9Rigorous Mathematical Foundations of Data Field TheoryIV. Geometric structure
CH10Advanced Extensions of Data Field TheoryIV. Geometric structure
CH11From Theory to Computation: The DFT SolverV. Numerical methods
CH12DFT Learning Stack: From PDEs to Trainable SystemsV. Numerical methods
CH13Critical Phenomena in the Wild: Universal DFT Signatures Across DomainsVI. Empirics
CH14Non-Equilibrium Learning: Parrondo Switching in PracticeVI. Empirics
CH15From Principles to Product: The Data Field Processor (DFP)VII. Implementation
CH16Geometry and Topology of Concepts: Hyperbolic Embeddings and Gauge-Theoretic ExtensionsVIII. Applications
CH17Agents as Fields: DFT for Reinforcement Learning and PlanningVIII. Applications
CH18Conclusion and Synthesis: The Data Field Theory of IntelligenceIX. Synthesis
CH19Integrating DFT with Large-Scale Foundation Models: Geometric Priors, Spectral RegularizationX. Foundation models
CH20DFT-Native World Models: Architectures, Benchmarks, and ScalingXI. Agents and environments
CH21Field-Theoretic Agents for General-Scope Tasks: From Continuous Control to AGIXI. Agents and environments
CH22Scaling Geometric and Topological Concept Stacks: End-to-End Pipelines for Web-Scale KnowledgeXII. Topological aspects
CH23DFT as a Safety and Alignment Layer: Diagnostic Monitoring and Control for AGI SystemsXIII. Safety
CH24Full-Stack DFT Systems with Hardware Co-Design: End-to-End AGI PlatformsXIV. Integration
CH25Empirical Validation of DFT Predictions at Scale: Large-Scale Experiments and Reproducibility KitsXV. Validation

Computational framework

dft-solver

A Python implementation of the numerical machinery of DFT: spectral, finite-element, and graph-Laplacian discretizations; energy-stable IMEX time integration with discrete dF/dt ≤ 0 guarantee; adaptive timestep respecting finite-speed CFL; and an MMS + convergence + spectral + boundary-condition validation harness.

The solver is under active development. Distribution is currently limited; access is available for academic collaboration on request.

dft-solver v1.0
─────────────────────────────────────
discretization   spectral · FEM · graph
time integration IMEX · adaptive · symplectic
operators        Δ_g · ∇_g · GL[m², λ]
validation       MMS · convergence · spectral · BC
diagnostics      m_gap · ξ · v_c · 1/f^β

03

FEATURED RESULT

Non-equilibrium Parrondo switching.

Two training regimes, each individually counterproductive, can be alternated at the right frequency to produce a net improvement in the loss function. The phenomenon, borrowed from the Parrondo paradox in stochastic dynamics, is a first-of-its-kind application to learning. The optimal switching rate is derived from a Floquet / joint-spectral-radius analysis of the combined operator.

training step (t) loss / energy A alone ↑ B alone ↑ A ⇄ B ↓

Non-equilibrium Parrondo switching. Two regimes (orange, blue) each increase loss when run alone. Alternating between them at the Floquet-resonant switching frequency (dashed vertical marks) produces a net decrease, the joint spectral radius of the alternation falls below unity even when each regime individually has spectral radius above unity. The Parrondo-switching chapter develops the spectral diagnostics; the hierarchical-Parrondo paper extends to multi-scale stacks.

04

PATENT DRAFT

Patent draft: 28 claims.

The patent draft covers six categories: the core field-theoretic learning method (claims 1-9), discretization and solver techniques (10-12), the learning stack with mass-gap regularization, RG schedules, and Parrondo switching (17-21), agents and world models (13-14, 26), foundation models (23), and the safety / alignment layer (27).

Filing strategy, provisional vs non-provisional, jurisdictions, continuations, is being finalized. Coverage matrix maps each claim to the chapters and papers that develop it.