WFGY 2.0. Semantic Reasoning Engine for LLMs (MIT). Fixes RAG/OCR drift, collapse & "ghost matches" via symbolic overlays + logic patches. Autoboot; OneLine & Flagship. ⭐ Star if you explore semantic RAG or hallucination mitigation.

• Updated Oct 14, 2025
• Python

AI Safety research platform for studying personality drift in AI systems using mechanistic interpretability and clinical assessment tools. Complete simulation framework with neural circuit analysis, statistical drift detection, and intervention protocols.

• Updated Aug 1, 2025
• Jupyter Notebook

I Asked It to Forget, but It Didn't — A Case of Miscommunication Between AI and Humans

• Updated Apr 17, 2025

📦 Redwood Research's transformer interpretability tools, conveniently packaged in a Docker container for simple and reproducible deployments.

• Updated Apr 21, 2024
• Dockerfile

A Recursive Lab for Visual Intelligence - A framework for applying structural pressure, interpretability reasoning, quality assessment and constraint-based analysis to expose image collapse and machine vision failure modes - and rebuilding images outside of engine defaults – No AI training, dataset scraping, or derivative generation permitted.

• Updated Oct 17, 2025
• Jupyter Notebook