Original research by Kristina Leticia Rodriguez.
A Soul Matrix is a method for developing an AI-inhabitable synthetic person's psychological architecture, built through empirical prompt engineering of a structured constraint document grounded in wound-based causality. The approach is designed to be model-agnostic, portable, scalable, and customizable.
Unlike outside-in personality specifications that enumerate traits or if-then behavioral rules, the Soul Matrix is built inside-out: real-life wound experiences form the foundation, and those wounds refract across the architecture to supply the causality behind each trait. Contradictions between encoded desires and encoded aversions (polarity pairs) are preserved rather than resolved, and emergent in-character behavior appears to arise from the model navigating those contradictions.
- What is a Soul Matrix (Psychological Architecture) - Overview of the methodology, its load-bearing sections, the role of contradiction in emergent behavior, and early cross-model observations.
- Soul Matrix Framework v1.0.0 - Blank architecture template.
- Soul Matrix System Prompt & Post History Instructions v1.0.0 - Blank template for the operational document that tells a model how to drive an architecture.
The two blank frameworks are templates to be filled in. For what each section does and why it is structured that way, see the methodology document above.
Early-stage independent research, 2026. Observations to date span multiple model architectures including Kimi K2.5, Kimi K2 0905, GLM 5.1, DeepSeek V3 0324 finetune, DeepSeek V4 finetune, Claude Opus 4.5, and Claude Fable 5. Findings are promising but preliminary; no formal claims are made pending further documentation and trials.
Soul Matrices are authored human-in-the-loop. Every section of the framework is written by the author: personality traits, wounds, desires (secretly loves), polarity pairs, background, family, home, fashion, likes, physical tells, and all remaining fields. The one-sentence causal expansions attached to each trait are generated with LLM assistance, constrained to derive from the author-written sections, and reviewed and revised by the author. The character System Prompt and Post History Instructions are generated with LLM assistance from author-specified requirements, then reviewed, edited, or rejected by the author. Whether model-generated causality is easier for models to navigate than human-written causality is an open question and a live limitation.
Character specifications for the author's own universe are held private; the methodology is public. Royston Thorne exists in both: a private full version and this standalone, setting-neutral demo built for public release.
See Demo - Royston Thorne SOUL MATRIX - v1.0.0.md for the worked example.
Active Independent Research, 2026. Feedback welcome.
All content in this repository is the original work of Kristina Leticia Rodriguez, (c) 2026.
Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).
You may read and share this work with credit. You may not use it commercially or distribute modified versions without permission.
Full license terms: https://creativecommons.org/licenses/by-nc-nd/4.0/
If you reference this work, please cite: Rodriguez, K. L. (2026). Soul Matrix Methodology. GitHub repository.
For collaboration or permissions inquiries, contact: KLRod47@gmail.com
This work is provided as-is, without warranty of any kind. It is research material, not a product. Royston Thorne is a fictional character built for demonstration and testing; he is not a licensed clinician and is not a substitute for professional mental health care or for a real relationship. The author is not responsible for how this material is used or for any outcomes resulting from its use.