Composers have spent more than fifty years devising computer programs for the semi-automated production of music. This article shall focus in particular on the case of minimal run-time human intervention, where a program allows the creation of a musical variation, typically unravelling in realtime, on demand. These systems have the capacity to vary their output with each run, often from no more input information than the seeding of a random number generator with the start time. Such artworks are accumulating, released online as downloads, or exhibited through streaming radio sites such as rand()%. Listener/users and composer/designers may wish for deeper insight into these programs' ontological status, mechanisms and creative potential. These works are challenging to dissect; this article makes a tentative start at confronting the unique problems and rich behaviours of computer-program-based generative music, from the social and historical context to the backwards engineering of programs in relation to their sound world. After a discussion of exemplars and definitions of generative art, strategies for analysis are outlined. To provide practical examples, analyses are provided of two small scale works by James McCartney.