Using Agentic Reinforcement Learning to Play Pong