Last updated: March 18, 2026
Singularity Cybersecurity
Building Singularity Cybersecurity — an AI security research lab. Three active experiments:
AgentArmor [HYPOTHESIZED] — runtime behavioral monitoring for AI agents. Experiment #1. SkillVet [HYPOTHESIZED] — behavioral sandbox scanning for third-party agent skills. Experiment #2. RedClaw [HYPOTHESIZED] — automated red-team toolkit with OWASP Agentic mapping. Experiment #3.
Research
RL Agent Vulnerability — 40 agents, 4 attack classes, OWASP 7/10 mapped. Complete. Model Behavioral Fingerprinting — 30-combination benchmark. “Antivirus for AI models.” Adversarial Training Optimization — robustness-utility tradeoff for LLM safety. govML — 50+ templates, 10 profiles, 20+ generators. Methodology
Learning
MSCS Georgia Tech — Machine Learning specialization. 4/10 courses complete.
Reading
Thinking in Systems — Donella Meadows Anthropic research on AI safety evaluations OWASP Agentic Security Top 10