Given a succinct programming language designed for sound, it may be no surprise that people can write short, interesting music-generating programs. An album with 22 tracks by different artist/programmers was created in SuperCollider; each program fits in 140 characters. The code featured in this video is different in that it is written in general-purpose programming languages that do not have frameworks for music built in.
The first two posts explain more about this practice, one that arises in the context of the demoscene, where programmers keep and use knowledge of platforms and (often) assembly-language programming to produce pleasing audiovisuals. There are connections to various sorts of "recreational computing" and to one-line programming in BASIC and APL.
But, consider the presentation of this code on YouTube. For all the earnest, explanatory drive that Viznut exhibits - providing a visualization, offering the code itself and textual comments - there's something absurd about this media object. An extremely simple computer program is being viewed after being rendered as a video and delivered in a browser using Flash, with time, effort, and computation spent on both ends. Instead of tiny, flexible programs, we have a huge streaming video, difficult to download and very limited. No modification, no freedom to run a program indefinitely.
Of course, there are ways to play around with code, to modify and write new sound-generating programs, even without compiling C, as the third and fourth links show - links that would not be visible when viewing the embedded video by itself.
YouTube is ascendant. We lack YouCode. Even if you can tweet SuperCollider programs, the prevalent, popular systems today are designed to show controlled media fragments and to keep the user from using her computer as a computer in the way that Viznut and other cultural workers are so effectively doing. In Media Res, which allows videos and images but not running computer programs as objects of discussion, could innovate here.