The recent advancements in the field of in-memory computing (IMC) and neural network optimization for low-power applications are significantly reshaping the landscape of modern computing. Researchers are increasingly focusing on leveraging novel memory technologies, such as Y-Flash, to enhance the efficiency and performance of machine learning inference, particularly for large-scale data processing. This approach not only addresses the limitations of traditional von Neumann architectures but also introduces new possibilities for energy-efficient, high-accuracy inference models. Additionally, the integration of neural networks with printed electronics is pushing the boundaries of ultra-low-power, flexible computing, enabling the deployment of complex models in wearable and implantable devices. These developments are paving the way for more efficient and scalable solutions in the realm of edge computing and IoT applications. Furthermore, the training of physical neural networks for analog IMC is addressing the challenges posed by hardware non-idealities, offering a promising direction for improving the accuracy and reliability of IMC systems. Overall, the field is moving towards more integrated, energy-efficient, and flexible computing solutions that cater to the demands of modern, data-intensive applications.