Planning now names the decisions the agent used to guess at. It asks which image style you want and holds that style across every image in the lesson, and when documents are attached, it asks how each one should be used — reproduced closely, restructured into an interactive lesson, or treated as background to build beyond. It also hands you the call on when planning ends, rather than declaring itself ready.
Video works the same way. Before authoring a single scene, the agent settles the question that shapes the whole video and the spend behind it: an on-camera presenter or narrated motion graphics, plus the visual style for the scenes.
It only asks about what is genuinely still open. Choices you already made in the create menu, the lesson title, an approved storyboard, or your own message are treated as settled, and language, voice, and styling always come from the lesson defaults. In a course it asks once for the whole run.