SysAdmin: Measuring Instrumental Power-Seeking in Frontier AI Enhancing Rubric-based RL via Self-Distillation SFGA: A Statistics-First Gating Architecture with Adjudicative Escalation for Trustworthy SFT Data Procurement Dual Adversarial Fine-tuning for Enhancing Robustness of Large Vision Language Model Alignment of a Total Automation Economy