I am a researcher bridging the gap between Information Geometry and Deep Learning. My work focuses on rigorous mathematical proofs for AI stability in recursive training loops.

Currently, I am investigating Model Autophagy Disorder (MAD)—the topological collapse of AI models trained on synthetic data—and proposing Salmon Regularization as a counter-entropic solution.

• 🔭 Current Research: The Ainex Singularity (Geometric Proof of Dimensional Collapse).
• 🏛️ Affiliation: King Abdulaziz & His Companions Foundation for Giftedness and Creativity (Mawhiba).
• ⚡ Focus: Moving beyond "Scaling Laws" to "Geometric Grounding."

Submitted to Journal of Artificial Intelligence Research (JAIR) & UAI 2026.

We formally prove that static neural topologies impose a "Rigidity Penalty," guaranteeing that recursive loops contract the semantic convex hull to a zero-information singularity.

Status	Venue	Artifacts
Under Review	JAIR	[Manuscript Submitted]
Under Review	UAI 2026	[OpenReview]
Preprint	Zenodo
Code	GitHub	Ainex-Limit-Experiment

I maintain a connected network of resources to document the research journey and share tools:

• Hugging Face: Hosting the Ainex-GPT2 checkpoints and collapse visualizations.
• GitHub: Open-source implementation of Salmon Regularization.

• Dev.to: Breakdowns of complex entropy concepts for engineers.
• CoderLegions: Developer-focused tutorials and discussions.
• HackerNews: Discussions on AI safety and recursive training risks.

• Core: PyTorch, NumPy, SciPy, LaTeX.
• Analysis: Information Thermodynamics, Topology, Differential Geometry.
• Workflow: Git, Docker, VS Code, Linux.