The field of robotics and simulation is rapidly advancing, with a focus on creating more realistic and scalable environments for training and testing autonomous agents. Recent developments have centered around the use of data-driven approaches to generate high-fidelity simulation scenes, allowing for more accurate and generalizable training of robotic policies. Notable advancements include the use of 3D Gaussian Splatting for photorealistic rendering and the development of novel frameworks for sim-to-real transfer and robotic manipulation. These innovations have the potential to significantly improve the performance and adaptability of autonomous robots in real-world environments. Noteworthy papers include: UrbanVerse, which introduces a data-driven real-to-sim system for converting city-tour videos into interactive simulation scenes, and GaussGym, which presents an open-source framework for learning locomotion from pixels using 3D Gaussian Splatting. GRASPLAT is also notable for its novel approach to dexterous grasping through novel view synthesis. RubbleSim and GSWorld are also highlighted for their contributions to photorealistic simulation and robotic manipulation.