I've pulled all-nighters for star parties. Takes a while to get back on schedule, but it's been worth it.<p>However, the thesis is that an all-nighter is bad for the project. Unfortunately, the author mixes up an occasional all-nighter ("just this once") with frequent all-nighters ("If you are at a point where you are working 24 hours a day"), and without making a distinction between the two.<p>The problem with this is that I personally can point to cases where an all-nighter worked - we added a couple of new features leading up to a demo at a supercomputing conference, and stayed up all night the night before leaving. They worked, and we scheduled in time to clean up the code.<p>The author calls out to "design" (by which I assume he means evolutionarily designed), but omits cases like the story of Cliff Young, and the influence on all-nighters for ultra-marathons. Clearly the best solution in that case is to not sleep, and I will argue there are evolutionary reasons why we <i>can</i> do that.