[_mql]

It's meant to be very strict about semantics (LaTeX actually isn't - it has a lot of style-related commands), and it will be extensible (so you can add your own content types). It's based based on a JSON document model that is describing a series of operations that can be used to reconstruct any document state and also enabling realtime collaboration. Authoring will take place in WYSIWYG-fashion. By having a real data-representation of the document, it can be turned into anything (e.g. LaTeX, PDF, HTML).

We use Pandoc to support a variety of output formats already.

E.g. goto: http://substance.io/michael/data-js and open the export dialogue.

However, we're working on a new architecture (https://github.com/substance/architecture) because we don't want to treat the symptoms of the current implementation and instead fix the problems we've identified at heart.