Researchers have introduced RANGER, a novel approach to pathology report generation that leverages sparsely-gated mixture-of-experts with adaptive retrieval re-ranking. This method addresses the limitations of existing transformer-based architectures, which often rely on homogeneous decoder structures and static knowledge retrieval integration. By incorporating adaptive retrieval re-ranking, RANGER can better handle the complex morphological heterogeneity of Whole Slide Images. The use of sparsely-gated mixture-of-experts enables more efficient processing of large-scale images, making it a significant advancement in the field of pathology report generation1. This development has significant implications for the field of medical imaging, as it can improve the accuracy and efficiency of pathology report generation. The shift towards more sophisticated architectures also reflects a broader trend in AI research, where state-aligned activity is driving innovation and changing the threat model. So what matters to practitioners is that RANGER's innovative approach can potentially raise the bar for geopolitical actors seeking to exploit AI vulnerabilities.