The field of reinforcement learning is moving towards improving stability and optimality in learning processes. Researchers are exploring new methods to mitigate oscillations and overestimation bias, leading to more efficient and effective algorithms. Notably, innovative approaches are being developed to enhance algorithmic stability, such as selectively updating policies and employing adaptive adjustment mechanisms. Additionally, there is a growing interest in performative reinforcement learning, which aims to optimize policies that can change environmental dynamics. Recent works have made significant progress in achieving performatively optimal policies, which maximize the original value function. Overall, the field is advancing towards more stable, optimal, and effective reinforcement learning algorithms. Noteworthy papers include: Use of target networks is being revisited with the introduction of a novel update rule, allowing for faster and more stable value function learning. A zeroth-order Frank-Wolfe algorithm has been proposed to achieve performatively optimal policies in polynomial time, under certain conditions. Analysis of credal set updates has provided insights into the conditions under which stability emerges in iterative learning processes.