I specialize in discovering "Invisible Rules" within neural networks. My work focuses on Mechanistic Interpretability and Neuro-Symbolic architectures, moving away from massive data dependency to focus on pure logical emergence.

• C.L.E.A.R. Framework: A subtractive AI architecture that extracts minimal logical kernels (Diamants Logiques) through high-pressure weight decay.
• Axiomatic Duality: My research on the "Quantum Unifier," demonstrating how minimalist networks can resolve direct logical paradoxes by discovering latent wave-functions.
• OpenAura Search: Implementing advanced search and logic-driven retrieval systems.

• Grokking & Phase Transitions: Studying the sudden shift from memorization to algorithmic understanding.
• Subtractive Intelligence: Developing the "Sieve" (Passoire) methodology to eliminate statistical noise in favor of pure logic.
• Logic Transplantation: Grafting independent reasoning kernels into universal master cores.

I am open to collaborations on Mechanistic Interpretability, Grokking research, or any project pushing the boundaries of non-probabilistic AI.

• Email: ronoxdev.contact@gmail.com
• X (Twitter): Maxenonyme