From small soft machines to muscle-powered pumps, biohybrid robots are expanding possibilities in medicine and engineering. In a review for the International Journal of Extreme Manufacturing, Dr. Su Ryon Shin and her team at Harvard Medical School highlight advances in biofabrication, from 3D bioprinting to microfluidics that integrate living muscle cells with robotic systems. Their roadmap outlines how these technologies could move biohybrid robotics from lab concepts to real-world tools for regenerative medicine, drug testing, and disease modeling.

Introducing Genos, an open genomic foundation model for high-quality human genomes across global populations, built to advance understanding of the genome and its diversity. Genos achieves single-nucleotide precision over a context of 1 million bases and uses a Mixture of Experts (MoE) architecture that is practical to deploy and economical to run. We provide two variants, 10B and 1.2B, so teams can match compute and latency needs. Trained on 636 high-quality assemblies representing diverse ancestries, with ~1.6T training tokens for the 1.2B model and ~2.2T for the 10B model, Genos learns the genome itself and generalizes to population-scale analyses. With efficient routing and scalable inference, Genos connects sequence to function and powers tools for discovery and diagnostics, and a complementary 4B multimodal language model pairs genomic signals with text to enable omics-aware reasoning in those tools. Explore code, weights, and docs on GitHub, Hugging Face, ModelScope, and zero2x.