Building an infinitely learning geolocation game with RL | Dark Hacker News