I think to a good degree experience in the subject matter also comes into play with this. If your bread and butter expertise lies in a certain area and you were

I think to a good degree experience in the subject matter also comes into play with this. If your bread and butter expertise lies in a certain area and you were asked to write a small self-contained application in that area from scratch, you probably end up with something much closer to the goal than the white triangle on a black screen. I found it very enlightening to sit through the almost full 48 hours of Notch programming some Ludum Dare entries which he streamed on twitch.tv. The end product is often very hackish, but that's exactly what this article argues could be a strategy to success. In any event, you can tell that he has programmed similar exercises many, many times in his life because certain things that would take me days to get right, he just writes down more or less without thinking. Not writing your own engine is also a recommendation that can be seen from two sides. Of course, for a lot of us, it is tempting to go and try writing a re-usable, nicely abstracted and modular piece of software. But for a 48h jam, that can often be the wrong direction, because you get side-tracked writing your nice game engine and meanwhile forget to actually write the game. However, at the same time, referring back to the aforementioned videos by Notch, he likes writing everything from scratch and not use a pre-existing engine because it can give you full control over what you can do and how, which for a 48-hour jam might be what you want. (Notch and these videos have been the subject of discussion on HN many times before, so there's probably no point to going down that road here again. I personally find it sad that he doesn't seem to stream any more programming vids.)