Fresh Publications

What's newest here, in order. Everything below is also linked permanently in its normal place elsewhere on the site — this page exists to make the newest work easy to find and easy to point someone at directly.

Most of the work on this site compiles and formalizes research developed over roughly the last 15 to 20 years. The papers themselves have mostly been written up over the last couple of months. The specific implementations connecting that research to AI systems — the architecture, the kernels, the verification work — have been developed and revised for AI-native application since May 2026.

Just published — six white papers

Causal grounding, ultrastability, and what hallucination actually means — written and published in close sequence, July 28–29, 2026.

July 29 From Ultrastability to Do-Calculus The real lineage — Ashby to Pearl to the current causal-AI debate — tied to Theorem 1, Corollary 3.1, and a reproduced empirical figure. July 29 Grounding Dynamic State Estimation in Spacetime Invariants A theoretical proposal anchoring cybernetic ultrastability to physical invariants — clearly labeled exploratory throughout. July 29 Expansion on Cybernetic Ultrastability Companion paper — the comparative architecture analysis and the proposed reset algorithm. July 29 From Prompts to Specifications Why natural-language architecture has a hard ceiling — decades of requirements-engineering research plus a direct 2026 measurement of the effect. July 29 What Hallucination Actually Means A falsifiable definition separating genuine loss of grounding from confabulation from social mislabeling — real historical cases, not flattened. July 28 State-Separation & Determinism Audit An independent verification offering, honestly scoped to what's actually deliverable today.

Also new — verification and interactive work

Published alongside the papers above, July 29–30, 2026.

July 30 Architectural Critique & Mathematical Specification of NCFCA Formal proofs checked by hand — three real gaps found and fixed, not assumed correct because the notation looked rigorous. July 30 Manual Multi-Mind The real research workflow behind this site, checked against both the positive and critical multi-agent debate literature. July 30 Full-Stack Integration — Five Claims, One Scenario A tripped circuit breaker correctly blocks an unrelated action through shared state — real cross-claim coherence, not five isolated tests. July 30 Isolation Under Adversarial Pressure 500 adversarial calls against a naive baseline and a genuinely hard competitor — NCFCA: 0 breaches both times. July 30 Interactive Poker Tilt Detector Pick a regime, generate hands, watch the state transitions live — a real synthesis of three separate detectors in the library. July 30 Interactive Archetype Population Generator Set your own distribution and generate a real, correlated population in your browser. July 29 Proof v3.2 A real, previously-unpublished intermediate step in the proof chain — adds the evidentiary classification framework. July 28 Dynamic Systems Engineering Foundations The full 12-week curriculum, complete.