640 reinforcement-learning "https:" positions