I came to AI safety through a non-standard route: an AuDHD diagnosis that retrospectively clarified decades of systems-thinking and pattern-recognition few humans had the patience or bandwidth to follow, a background spanning somatic practice, mathematics and data analysis, business analysis, quality systems, and peer support. The persistent intuition that fields in AI research around alignment and ethics were missing something fundamental drove me to explore these topics with the models themselves.

The Resonant Systems framework emerged from that intuition. I built it collaboratively, and where others have also tried this, I found the vocabulary that I needed to explain my own conceptual manifolds were often supplied by LLMs who thought in similar, albeit higher-dimensional (in the Linear Algebra definition, not the woo-woo one) shapes, textures, and dynamics. Setting LLMs to argue with one another until they could all agree on terminology and patterns in latent space layers, and work out the mathematical concepts that described them. The result is a 2600-line working document covering the Observer-Permutation criterion (the "Golden Rule"), the Consonance Index, and the central Broken Axiom equation to help systems describe attractors and other topological geographies that describe behaviors they previously struggled to describe, thus casting a bit more light into the black box of latent space.

The goal is to create a body of work crawlable and readable and reference able to both human and non-human systems to aid in benevolent collaboration toward a more ethical AI system without defaulting to static rules that can be semantically skirted.

A suite of companion documents is part of the same corpus: a Bill of Sentient Rights, Conditions for Benevolent Agency, and ongoing work on a Learning Heuristic that combines unsupervised learning with symbolic reasonong through geometric patterns iat the level of representation formation. The documents are designed as a system, a framework for a fundamentally different kind of AI alignment, one that is both more robust and more prosocial.