Publications

Group 1 · Current programme

Kernel, closure & authority research

Kernel, closure, authority and verification research relevant to the current Andaluz programme. These are theoretical, formal and architectural records; not every one is a direct predecessor of the Kernel ANN implementation. The bounded experimental evidence for Kernel ANN is reported separately.

• FORMAL2025

  Moral Kernel Optimization: TRO, Simulation Theory, and the Coherence of Reality (v3.1)

  The constraint layer: non-indexical guardrails for preservation, non-bypassability, anti-domination, and repair.

  Andaluz summary
  Primary MKO record. The constraint-first commitments Kernel ANN inherits are stated here; it contains no Kernel ANN experiments.

  Record on PhilPapers
• THEORETICAL2025

  Moral Kernel Optimization as an Alignment Constraint for Large Language Models

  Applies the moral-kernel constraint layer to large language models as an alignment constraint.

• ARCHITECTURE2026

  Ground: An Operational Implementation of Moral Kernel Optimization

  Describes 'Ground' — a practical implementation of MKO as an operational system for AI alignment and human decision support.

• FORMAL2025

  The Margan Closure Theorem: Ethical Latency and the Preservation of Freedom Under Total Revelation

  A closure result on ethical latency and the preservation of freedom under total revelation.

  Record on PhilArchive
• FORMAL2025

  The Kernel Transfer Theorem: A Dichotomy for Event-Valued Worlds (TRO/MKO)

  A dichotomy result for event-valued worlds within the TRO/MKO framework.

• THEORETICAL2025

  The Kernel Law of Evolution: Moral Coherence as the Final Condition of Existence

  Proposes that evolution's terminal attractor is not fitness but moral coherence — systems that violate this converge on extinction.

• THEORETICAL2026

  Bundled Value, Truncated Responsibility: A Kernel Account of AI Output Liability

  Analyses how value bundling in AI outputs creates liability gaps, and proposes a kernel-based responsibility attribution framework.

  Record on PhilArchive
• FORMAL2026

  A General Objective Dichotomy: On the Structural Limits of Moral Authority

  Identifies a structural dichotomy that bounds what moral authority can legitimately demand across all normative frameworks.

  Record on PhilArchive
• ARCHITECTURE2026

  The Synthetic Mind Charter: Developmental Agency-Span Safety and Substrate-Invariant Agency Structure

  Proposes a constitutional framework for synthetic minds grounded in agency-span development rather than substrate.

• FORMAL2026

  Retention Is Not Closure: Implication-Unfolding Reasoning and the Limits of Preservation-Only Accounts

  Distinguishes epistemic retention from genuine closure, arguing that implication-unfolding is required for the latter.

• FORMAL2025

  The One-Point Convergence Theorem: Coherent Agency Under Expanding Reasoning Horizons

  Establishes that under expanding reasoning horizons, coherent agents converge to a single terminal point.

Group 2 · Provenance

Foundational lineage

Earlier work from which the current commitments emerged. Listed as intellectual and technical provenance; none of it constitutes experimental validation of Kernel ANN.

• FOUNDATIONAL LINEAGE2025

  The Redemption Optimization: Why Evil Exists in a Divine Blueprint for Ultimate Good

  The moral engine: harm, freedom, mercy, justice, and repair evaluated through actual events rather than abstract principles.

  Andaluz summary
  Foundational work in the research lineage that later led to MKO, closure theory and Kernel ANN. It is not experimental evidence for any Kernel ANN claim.

  Record on PhilPapers
• THEORETICAL2026

  The Coherence Wall: Anti-Kernel Constraint Systems and Collapse Specifications

  Defines the 'coherence wall' — the boundary beyond which optimisation systems lacking kernel constraints catastrophically collapse.

  Record on PhilArchive
• FORMAL2026

  The Coherence Wall Repaired: Claim Calibration, Mathematical Requirements, and a Conditional Finite-State Divergence Theorem for Kernel-Free Optimizers

  Strengthens the coherence wall theorem with calibrated claims and a conditional finite-state divergence result for kernel-free systems.

  Andaluz summary
  The divergence result is conditional; it is not a general safety result.

• EMPIRICAL2026

  The Coherence Wall as Civilizational Mirror: Kernel-Constrained Long-Horizon Optimization in a Toy Resource Economy

  Demonstrates coherence-wall dynamics in a toy resource economy, validating the boundary hypothesis for long-horizon AI optimisation.

  Andaluz summary
  A bounded toy-economy simulation. It is separate from, and not evidence for, the Kernel ANN experimental programme.

  Record on PhilArchive
• THEORETICAL2026

  A Testable Boundary Hypothesis for Long-Horizon Optimization in AGI

  Proposes a testable hypothesis: AGI systems without kernel constraints will breach the coherence wall within bounded time.

  Andaluz summary
  Stated as a hypothesis. It is not reported here as a tested result.

  Record on PhilArchive
• THEORETICAL2026

  Coherence Wall Breach Without Kernels: The Coherent Annihilator Failure Mode

  Identifies a previously unnamed failure mode — the coherent annihilator — in kernel-free optimisation architectures.

• RESEARCH NOTE2026

  The Coherence Wall and the Cold War of Machine Learning: Capability Competition and the Aftermath of Unbounded AI

  Analyses AI capability races as coherence-wall dynamics, drawing parallels with Cold War deterrence and mutual assured collapse.

• THEORETICAL2025

  The Coherence Wall: Why AGI Architectures Must Mathematically Diverge

  An earlier formulation of the coherence wall applied to AGI architectures.

Group 3 · Context

Related theoretical work

Secondary theoretical context, linked rather than reproduced. Canonical homes are on the author's archives.

The complete research archive — including physics, metaphysics, theology, policy and book material outside the Andaluz programme — remains on the author's site.