The field of computer vision is moving towards more efficient and effective models for visual encoding and road network extraction. Recent developments have focused on exploring the representational potentials of new models and architectures, such as Mamba, and improving their performance on various tasks. Notably, there is a trend towards leveraging the strengths of different models and combining them to achieve state-of-the-art results. For instance, hybrid architectures that integrate sequential modeling with global reasoning have shown great promise in achieving topologically coherent road segmentation. Additionally, novel approaches to road network extraction, such as using differentiable Bezier graphs, have demonstrated impressive performance on large-scale benchmarks. Overall, the field is advancing rapidly, with a focus on developing more accurate, efficient, and interpretable models. Some noteworthy papers include: RNN as Linear Transformer, which provides valuable insights into the representational properties of Mamba models. DOGE, which introduces a new paradigm for generating high-fidelity vector maps of road networks. MambaEye, which proposes a size-agnostic visual encoder with causal sequential processing. PathMamba, which achieves topologically superior segmentation maps without prohibitive scaling costs.