With the advent of deep learning many of the tasks in computer science that have been deemed impossible suddenly became only a few clicks away. One of the approaches made available is reinforcement learning - a method for solving problems by establishing an action-reward scheme. Combined with the power and availability of the general-purpose game engines, anyone with a rudimentary knowledge of the topic can create and train their virtual creatures. In this talk we will use this power to solve one of the most frustratingly difficult (according to the internet) games of our era.