The reasoning engine approach is interesting - most video AI tools generate based on prompts but can't really understand or evaluate their own output quality. How does the Draft Mode actually work for iteration? Can you make specific changes to scenes without regenerating the entire video, or is it more about quickly testing different prompt variations? The HDR claims sound impressive but how does it perform with complex lighting scenarios? Most AI video still struggles with realistic shadows and reflections in multi-object scenes