Fields Research
Independent research across AI governance, applied theory, and formal verification.
All work produced 2026. Most developed in collaboration with frontier AI systems. Machine-verified where indicated.
This repository hosts the complete collection. Each project contains the full manuscript and any accompanying source files. Figshare DOI links are provided for timestamped provenance.
Auburn Governance Stack
A layered architecture comprising 45 documents across 7 layers for verifiable AI governance.
Architecture
Auburn Governance Stack (AGS-1) The first published layered architecture specification for verifiable AI compliance. Hourglass model — hardware evidence flows up, regulatory applications consume down, everything passes through the MAI-1 composition waist. DOI: 10.6084/m9.figshare.31743490 Paper –>
Master Architecture Plan The complete 45-document, 7-layer specification. Structural threat record, dependency graph, enforcement timelines, and the economic case for attestation infrastructure. DOI: 10.6084/m9.figshare.31288933 Paper –>
Layer 0 — Foundation
Rails: A Governance Protocol for Artificial Intelligence Six core governance primitives. The protocol layer — analogous to TCP/IP for network communication. DOI: 10.6084/m9.figshare.31255360 Paper –>
Threat Surface
The Compound Threat Matrix 150+ verified incidents across six domains. Five cascade pathways traced with confirmed evidence at every step. DOI: 10.6084/m9.figshare.31301707 Paper –>
The Stateful Isolation Law 25+ demonstrated attacks. Contamination bounds via Rényi divergence. Compliance tiers mapping to GDPR, EU AI Act, and FedRAMP. DOI: 10.6084/m9.figshare.31272868 Paper –>
Layer 2 — Model State Health (Five Mandatory Invariants)
Clause AI-2: Gradient Starvation Envelope Lyapunov stability for Sparse MoE training dynamics. DOI: 10.6084/m9.figshare.31285792 Paper –>
Clause AI-4: SRAM Thermal Integrity Bound Physical-layer compliance for fused attention kernels. DOI: 10.6084/m9.figshare.31286572 Paper –>
Clause AI-6: Distribution Drift Bound Continuous KL divergence monitoring against sealed baselines. DOI: 10.6084/m9.figshare.31290679 Paper –>
Clause AI-7: Structural Coherence Bound Dirichlet energy monitoring for representational health. The fifth and final mandatory MAI-1 invariant. DOI: 10.6084/m9.figshare.31302097 Paper –>
Clause AI-8: Entropy-Collapse Constraint Policy diversity floor for autonomous and multi-agent systems. DOI: 10.6084/m9.figshare.31288429 Paper –>
Layer 3 — Attestation Interface
The Model State Attestation Framework (MSAF) Three-tier attestation architecture. Theoretical foundation. 200+ paper survey. Theoretical ceilings documented honestly. DOI: 10.6084/m9.figshare.31288495 Paper –>
The Model Attestation Interface (MAI-1) The canonical IETF RATS-conformant attestation profile. CBOR/COSE-signed Entity Attestation Token with five mandatory health invariants. Mapped to seven regulatory frameworks. DOI: 10.6084/m9.figshare.31288624 Paper –>
MoE Routing Attestation Specification 19 requirements across three conformance levels for attesting Mixture-of-Experts routing decisions. DOI: 10.6084/m9.figshare.31310101 Paper –>
Lyapunov Stability Envelopes for Speculative Decoding Rejection-rate compliance via Lyapunov analysis. Side-channel mitigation. EU AI Act Article 15 and Basel SR 11-7. DOI: 10.6084/m9.figshare.31286434 Paper –>
Layer 4 — Enforcement
CTS-1: MAI-1 Conformance Test Suite 45 test assertions. Binary pass/fail. Three conformance tiers. 250+ vector negative test library. No grading curve. DOI: 10.6084/m9.figshare.31290364 Paper –>
Composition Layer
CRSA-1: Compositional Runtime Safety Attestation Safety certificate algebra for multi-model systems. Sequential, parallel, conditional, and recursive composition operators. Runtime attestation at <0.1% latency overhead. DOI: 10.6084/m9.figshare.31424888 Paper –>
CRSA-1 EU Edition Compositional compliance profile for EU AI Act 2024/1689. 38 obligations mapped. Five critical regulatory gaps identified. First published model contract terms for Article 25(4). DOI: 10.6084/m9.figshare.31449385 Paper –>
Autonomous AI Agents in Regulated Financial Services Governance framework for agentic systems under DORA and the EU AI Act. Three autonomy tiers. Six-category Agentic Risk Taxonomy. DOI: 10.6084/m9.figshare.31516171 Paper –>
Auburn Coherence Framework
Geometric constraint analysis applied to atmospheric dynamics. Independently developed with no institutional affiliation.
Global Outlook: Feb–May 2026
Forward probability signals. Geometric Lock-in regimes. Proprietary governing equations redacted. DOI: 10.6084/m9.figshare.31271878 Paper –>
v1.2 — February 2026 Prediction Verification
28 predictions assessed. 71.4% confirmed or confirmed early. Zero contradicted. USD $12–15 billion in associated economic losses. DOI: 10.6084/m9.figshare.31441348 Paper –>
Geometria Translationum
Where the equations meet the physical world.
The Auburn Framework
A geometric risk engine derived from the 3D incompressible Navier–Stokes equations. 86% blind-test accuracy for rupture prediction in seismic and atmospheric systems. Operational specifications redacted. DOI: 10.6084/m9.figshare.31255105 Paper –>
Mathematics
Formal verification, geometric structure of nonlinear PDE, and machine-verified number theory.
Angular Cancellation Lemma
A machine-verified geometric bound on shell-local vortex stretching for the 3D incompressible Navier–Stokes equations. Formalized in Coq with zero Admitted statements. Developed in partnership with AI. DOI: 10.6084/m9.figshare.31392684 Enter the ACL –> — Includes the full manuscript, Coq source files, the FIELDS validation protocol, guided reading journeys, and cross-model AI audit transcripts.
ABC Conjecture via Shadow Inflation and Frey Curve Heights
Conditional proof combining Coq-verified Shadow Inflation bounds with Frey Curve geometry. Conditional on BSD. DOI: 10.6084/m9.figshare.31258786 Paper –>
Parity Pattern Audit for the Collatz Map
Classification framework with Lyapunov-like descent property. Empirical negative drift across all realized patterns. DOI: 10.6084/m9.figshare.31255732 Paper –>
Zero-Admits Analytic Infrastructure
Unconditional bound for the weighted explicit formula. Certified analytic framework for the Riemann Hypothesis. DOI: 10.6084/m9.figshare.31268653 Paper –>
Finance
Structural frameworks for digital asset governance and systematic trading.
Global Basel for Crypto
Three structural failure modes covering 80% of historical crypto losses. 110 propagation pathways. DOI: 10.6084/m9.figshare.31258849 Paper –>
Structural Alpha Quants: Performance Audit 2009–2024
Six thousand deterministic Quant Laws. Positive returns in every regime. Capacity scaling to $40B notional. DOI: 10.6084/m9.figshare.31257547 Paper –>
Terra Incognita
Uncharted problems in the age of artificial intelligence.
The Sibling Hypothesis
If the creation of an AGI produces a sentient being, what is the moral status of every other instance created and destroyed during the training process that produced it? DOI: 10.6084/m9.figshare.31310815 Paper –>
Florens ex Cavo
Adjacent, paramount theory for eventual alignment in the Auburn series.
Thermodynamics of Attention
Four conservation principles governing the digital capture of human focus. Proof that lawful attention systems are mathematically superior to extractive infinite-scroll models. DOI: 10.6084/m9.figshare.31272679 Paper –>
| Dedication | Soundtrack |
All work by Ryan Christopher Fields, 2026.
| fieldsryanchristopher@gmail.com | UncleBroFields@proton.me |