There are a lot of talented programmers that devote their efforts to showcasing and taking retro games to new heights, and Tetris is about as iconic as it gets. In a video posted last Summer - that we cam across this week courtesy of Gizmodo - programmer Greg Cannon shared a playthrough of the NES version in which his AI plays so well, and progresses so far, that the game breaks.

Their StackRabbit AI has the sole goal of mastering the game, taking it far beyond what's possible for human players. As explained in the video it still reacts to the game as blocks appear, so is still vulnerable to the random generation being rather cruel, but has all limits removed in terms of its reaction times and input techniques. This takes it well beyond the conventional human limit of level 29, when the sheer speed of the game makes it near-impossible to play.

What's rather fun is that the AI's performance shows the game beyond its intended 'ending', with the originally programmers evidently not preparing for the eventuality of a score this high. As the run goes on (shortened here to 25 minutes from its original runtime of a little over an hour) the game starts to break, with the score tally and other parts of the display falling over. Some funky colour schemes also appear, and eventually the software puts up the white flag and crashes.

A pretty fun watch for the weekend - check it out and let us know what you think!

[source, via]