Windows/Mac/Linux: Though it may not look very different from other (dubiously beneficial) brain-training games, Brain Workshop is special in that there's a growing body of evidence to suggest that it may actually succeed in making its players a little bit smarter.
Free, cross platform and open source, Brain Workshop is founded upon the "dual n-back", a tricky sort of memory game where the objective is to keep track of two unrelated sequences of events at once. In most versions of the game, that means paying attention to a spoken list of letters while simultaneously tracking the changing position of a block on a three-by-three grid. Every turn, the game provides a letter and grid position and, every turn, the player has to indicate whether these newest values match the ones given a fixed number of turns ago. The more you play, the higher that number gets. Sound tiring? It can be! But it can also be kind of energising. The dual n-back is, in that sense, the only brain exercise we've encountered that actually feels like conventional exercise.
If it were like other self-styled "brain games", dual n-back training would probably make you very skilled at dual n-back training and little else. What researchers have found, however, is that, with a little practice, the game can bring about significant improvements in a person's short-term or "working" memory (often abbreviated WM), a faculty the University of Michigan describes as "the ability to maintain information in an active, easily retrieved state, especially under conditions of distraction or interference". It's believed that this improvement in working memory, in turn, can boost "fluid" intelligence, a person's ability to solve unfamiliar problems independent of acquired knowledge.
How much practice does it take to see those effects? Surprisingly little, though it does seem the more you put in, the more you get out. Study participants are generally asked to perform the task about 20 minutes a day for a period of three or four weeks. While the cognitive benefits of n-back training aren't permanent (and thus require upkeep to maintain), research shows that, at least in schoolchildren, they can persist for up to three months after training has stopped. And, interestingly, new research indicates that single n-back training may be just as effective as the more complicated dual-task sort. (Brain Workshop makes it easy to switch between dual and single n-back modes should you find you prefer one to the other.)
Standard postmodern caveats apply. Your own mileage may vary, yada yada. But, even if the positive effects are in your head, that's where your brains are anyway, right?