We present and explore a simple idea for improving document layout on arbitrary devices of different resolutions and size. The key idea is to allow manifold representations of content: multiple versions of anything that might appear in a document, such as text, images, or even stylistic conventions. Content is then selected and formatted dynamically, on the fly, by a layout engine in order to best adapt to a given viewing situation. We propose a user interface for authoring and editing such manifold content, and sketch a few interesting new algorithms that make use of it.