A research team from The Hong Kong University of Science and Technology (HKUST) has developed a novel AI framework, the Generalist–Specialist Collaboration (GSCo), which enables a generalist foundation model and specialist models to collaborate on disease diagnosis. GSCo outperforms both standalone models in tasks such as medical imaging, visual question answering, and radiology report generation, while significantly reducing development costs of medical models to adapt to new clinical tasks. This marks a meaningful step toward bringing AI medical models into everyday clinical workflows. The study, titled "Towards generalizable AI in medicine via Generalist–Specialist Collaboration," was published in the prestigious academic journal Nature Biomedical Engineering.

How much load can a crumpled sheet bear, and does size matter? A new study of two-dimensional macromolecules, a class of ultrathin sheet-like materials, shows that at equal packing density smaller sheets carry higher compressive loads than larger ones. Coarse-grained simulations and compression tests on paper, aluminum foil, PDMS, and silicone rubber reveal a universal “smaller is stronger” rule rooted in denser internal ridge networks, offering a new design principle for lightweight, high-strength structural metamaterials.

CuAu alloys with tailored compositions were integrated onto copper foam by a simple one-step electrodeposition method. The optimized Cu_0.25Au_0.75 catalyst exhibited rather excellent performance for the electrocatalytic oxidation of 5-hydroxymethylfurfural (HMF) via a one-electron dehydrogenation pathway. The improved catalytic activity could be attributed to the alloying-induced charge redistribution that modulates the d-band centers, thereby steering surface adsorption behaviors to balance HMF adsorption and HMFCA desorption, ensuring active site availability for the selective production of 5-hydroxymethyl-2-furancarboxylic acid (HMFCA) and simultaneous anodic hydrogen evolution.

With the growing demand for marine exploration and environmental monitoring, underwater sensing technology faces severe challenges due to high humidity and the need for simultaneous detection of multiple physical parameters, such as temperature and pressure. Traditional underwater sensors typically rely on combinations of various functional materials to achieve dual-mode temperature–pressure detection, often requiring additional waterproof encapsulation, which complicates system integration. Achieving dual-mode temperature–pressure detection within a single material underwater remains particularly challenging. Herein, a waterproof dual-mode sensor based on Bi₂Se₃/carbon paper (CP) composite films is developed by electrochemically depositing pyramid-interface-structured Bi₂Se₃ layers on both sides of a flexible CP substrate. The film exhibits enhanced thermoelectric properties, with a power factor of 106.0 μW m⁻¹ K⁻² (9 times that of CP). Sensors constructed from vertically stacked pyramid-interface-structured films enable high-precision detection of temperature and pressure underwater, featuring a temperature-sensing response time of 0.9 s and a pressure sensitivity of 0.94% kPa⁻¹, while allowing synchronous decoupling of the two signals. Moreover, the sensor exhibits excellent hydrophobicity, with a contact angle of 143.7°, along with robust stability, overcoming key limitations for underwater applications. This advance in flexible underwater sensing technology offers a reliable strategy for underwater human–machine interaction and monitoring.

A new Perspective in Science Bulletin argues that AI tools for intracranial aneurysm detection should be evaluated beyond test-set accuracy. The authors propose a four-stage pathway covering lesion-level validation, reader performance, prospective implementation and post-deployment surveillance to assess whether AI can deliver real clinical benefit.

This review summarizes recent advances in wearable uric acid sensors, covering electrochemical and optical detection mechanisms, performance-enhancement strategies, device integration, and emerging applications. It highlights progress in sensitivity, selectivity, anti-fouling performance, self-calibration, and user comfort, while identifying persistent challenges in long-term stability, interference resistance, multimodal monitoring, clinical validation, data security, and commercialization. Future directions emphasize intelligent, sustainable, and personalized sensing platforms.

By sending quantum information through a storm, researchers at The University of the Witwatersrand in Johannesburg, South Africa, have shown that the information in light can be kept completely intact, even though the light itself was completely warped beyond recognition.