The recent advancements in the field of neurosymbolic AI have demonstrated a significant convergence of symbolic and neural approaches, aiming to harness the interpretability and logical constraints of symbolic AI with the scalability and flexibility of neural networks. This integration has led to innovative methods such as relational neurosymbolic Markov models, which blend differentiable sequential models with relational logical constraints, and unified neurosymbolic systems capable of performing transformations as both symbolic and neural computations. These developments not only enhance the performance of AI models in sequential problems but also ensure they meet necessary constraints for trustworthy deployment. Additionally, the formalization and verification of AI systems have gained traction, with notable work on verified invertible lexers and the formalization of pattern matching in AI compilers, contributing to the robustness and reliability of AI applications. The integration of foundation models with relational programming frameworks, exemplified by Vieira, showcases a promising direction for creating versatile and accurate AI systems that can handle diverse tasks across different modalities. Overall, the fusion of symbolic reasoning and neural computation is driving the field towards more interpretable, reliable, and versatile AI systems.