Robert Musil’s literary estate provides a huge amount of material for research into a ‘phenomenology of writing’. This article introduces an annotation system for the digital representation of the estate from the perspective of writing. The dynamic changes of the text become visible in three different scenarios, corresponding with three levels of the genetic representation of the text. A) The ‘Minor (or Small-Scale) Writing Scene’ on the microgenetic level, whose traces in the draft manuscript allow us to distinguish five stages of revision. B) The ‘Middle (or Medium-Scale) Writing Scene’ on the mesogenetic level, taking into account additional manuscripts which play a role for sketching or rewriting a particular chapter draft. The distinction between draft and note, further differentiated into six types of manuscripts, is fundamental to the modelling of Musil’s writing process. C) The ‘Major (or Large-Scale) Writing Scene’ on the macrogenetic level, i.e. the text versions in their chronological dimension, with 45 dated periods (1898-1942), and their content-based location in the complex structure of the Man without Properties novel project, including preparatory projects, novel parts, chapter complexes and chapter projects. Digitally transcribed as early as the 1980s, the corpus of the digital estate was published in proprietary formats in 1992 and 2009, using a corpus-specific annotation system. Since 2016, MUSIL ONLINE, an open-access edition based on XML/TEI, has been in preparation at the Robert Musil Institute of the University of Klagenfurt. The second part of the article presents its full annotation scheme for the three scenarios mentioned.