The field of molecular generation and design is rapidly evolving, with a focus on developing innovative methods for generating small molecules and therapeutic peptides. Recent developments have centered around the use of autoregressive models, diffusion models, and latent variable transformers to improve the efficiency and accuracy of molecular generation. These approaches have shown significant potential in accelerating the discovery of novel therapeutics by enabling the targeted creation of molecules with desired properties. Notably, the integration of geometric awareness and conditional generation capabilities has enhanced the performance of these models. Furthermore, the development of specialized libraries and frameworks has facilitated the evaluation and comparison of different methods, promoting a more systematic and data-driven approach to molecular design. Noteworthy papers include: InertialAR, which introduces a canonical tokenization method and geometric rotary positional encoding to achieve state-of-the-art performance in unconditional molecule generation. MolChord, which proposes a structure-sequence alignment approach for protein-guided drug design, demonstrating state-of-the-art performance on key evaluation metrics. STAR-VAE, which presents a scalable latent-variable framework for molecular generation, enabling conditional generation and efficient finetuning with limited property and activity data.