An set of preprints benchmarks GPT‑4 against GPT‑3.5 across professional exams, critiques GPT‑4’s logical reasoning limits via diverse puzzles, reviews ChatGPT’s capabilities and constraints, assesses GPT‑3’s feasibility as a public health collaborator, and introduces ROSGPT, a ROS2 package enabling natural language–driven human‑robot interaction.