Short, technical notes on Safe AI, evaluation methodology, synthetic-data fidelity, and lessons from the research process.