Tech & AI

Researchers have exposed critical gaps in how nuclear fission models work by testing 10,000 variations of key parameters. The findings reveal that small tweaks to model inputs dramatically shift predictions for neutron and radiation output—insight that matters for nuclear energy safety, weapons stockpile stewardship, and reactor design.EN

Researchers have cracked a long-standing mathematical puzzle about infinitely repeating patterns in geometric space, opening new pathways for understanding fluid dynamics and physical systems. The breakthrough could accelerate computational models used in engineering, physics simulations, and financial risk analysis—fields where precise geometric mapping directly impacts product design and decision-making.EN

Researchers used deep learning with next-generation CT scanners to predict how proton beams interact with patient tissue, achieving 0.26% accuracy. Better dose targeting could reduce side effects and treatment failures in cancer care, while lowering the computational burden for radiation oncology centers.EN

Researchers found that speeding up laser 3D printing of high-performance metals creates more defects that dramatically reduce fatigue strength—a critical problem for aerospace and energy industries. The discovery reveals a hard tradeoff: manufacturers can't simply scale up production without sacrificing reliability in components that must survive repeated stress cycles.EN

Researchers have developed a faster way to train AI models on edge devices—smartphones, IoT sensors, factory equipment—by transmitting learning data over wireless networks without knowing channel conditions. The advance could accelerate real-time AI deployments in manufacturing, healthcare, and autonomous systems where latency costs money.EN

A new paper tackles a persistent blind spot in cultural heritage digitization: organizations create detailed 3D models and databases but rarely document *how* they were made. Researchers outline a practical framework for capturing this 'paradata'—the decisions, methods, and reasoning that explain outputs—which matters because undocumented processes become irreproducible, unreliable, and harder to reuse or improve.EN

Researchers have developed a cheaper, faster way to create reading comprehension questions that works across languages without requiring massive training datasets. The method could help schools and edtech companies provide personalized learning at scale, especially in languages where AI training data is scarce.EN

A new study reveals that electronic gates in Stockholm's metro don't just collect fares—they manipulate riders into policing each other, shifting labor costs to commuters. As transit agencies worldwide adopt similar technology to boost revenue, the research exposes hidden social costs that policymakers and operators rarely discuss.EN

Researchers developed a computational tool that predicts how quickly drugs dissolve and penetrate tissue barriers, eliminating months of repetitive lab experiments. The breakthrough could accelerate pharmaceutical development and lower costs for companies testing hundreds of candidate compounds.EN

Researchers have dramatically improved hydrogen gas output from algae by immobilizing cells in beads and using a chemical inhibitor to redirect energy flows. The advance could make algae a practical source of clean hydrogen fuel, potentially offering a renewable alternative to fossil-based production at scale.EN

Researchers spent six years testing a nature-integrated playground hut that tells interactive audio stories rather than displaying screens. The finding: children engaged more deeply, returned repeatedly, and invented their own narratives—suggesting a new model for educational tech that abandons screens entirely.EN

A new analysis argues that 'paradata'—invisible records of how data was collected and processed—could become crucial for verifying AI systems and meeting transparency demands. Yet researchers warn the concept needs clear ethical guidelines, or it could paradoxically reduce accountability rather than improve it.EN

A new study warns that large language models, while democratizing access to academic publishing, may inadvertently standardize how research is written and what knowledge gets valued globally. The finding matters for universities, publishers, and researchers trying to diversify scholarship—AI tools could undermine those diversity efforts by reinforcing dominant language norms and existing biases.EN

Researchers have mapped the mathematical rules governing how tree-shaped data structures develop, revealing that their complexity grows far slower than previously thought. The findings could improve how engineers design algorithms for everything from databases to network systems, potentially making data processing faster and more predictable at scale.EN

Scientists testing a new plasma deposition technique found it can coat surfaces at industrial speeds but generates damaging molten droplets that pit coatings and melt underlying materials. The findings suggest the method needs refinement before manufacturers can adopt it for precision coating applications in semiconductors and advanced materials.EN

Engineers have cracked why 3D-printed superalloy components for gas turbines lose critical strength during manufacturing—and mapped out a heating schedule to fix it. The discovery matters because it could unlock cost savings and design flexibility for aerospace and power generation companies already betting on additive manufacturing for high-stakes engine components.EN

Researchers found that combining wave and wind energy at the same offshore location boosts grid reliability far more than either technology alone. The finding offers utilities a blueprint for maximizing existing marine infrastructure and reducing the need for costly backup power as renewable energy expands.EN

Researchers used Swedish health records to predict results of a major heart disease drug trial before results were published—and the predictions held up. The finding suggests companies and regulators can trust observational data from hospital systems as a faster, cheaper alternative to traditional clinical trials for some medical decisions.EN

Researchers have developed sharper mathematical tools to detect when unmeasured factors secretly skew medical study results. The advance matters to pharmaceutical companies, regulators, and health insurers because it exposes flawed causal claims in treatment research—potentially saving money by preventing investment in treatments that don't actually work as claimed.EN

A year-long study of a Swedish e-science network found that new collaborative digital platforms don't reduce gender inequality in research — they replicate it. The finding challenges assumptions that modern project-based structures automatically promote diversity and signals that tech solutions alone won't solve entrenched workplace disparities.EN

Scientists have developed a method to distinguish between sensor errors and cyberattacks on wireless glucose monitors used in artificial pancreas systems. The advance matters because these devices—used by millions with diabetes—face growing security risks, and false alarms could delay treatment while missing real threats.EN

Researchers found that voice recognition AI fails dramatically when authenticating identical twins, with accuracy dropping sharply even as sample lengths increase. The finding exposes a significant vulnerability in biometric security systems now widely deployed in banking, law enforcement, and access control—raising questions about whether current voice ID technology is reliable enough for high-stakes identification.EN

Researchers have improved on the word cloud—a decades-old text analysis tool—by adding semantic positioning and multiple visual signals to reveal patterns in climate change documents. The upgrade could help policymakers, researchers, and companies quickly identify emerging themes and language shifts in climate communications, compliance documents, or media coverage.EN

Researchers tracked 30,000 people and found that chronic exposure to fine air particles accelerates coronary atherosclerosis—the dangerous buildup of plaque in heart arteries. The findings, based on actual imaging data rather than health claims alone, could reshape how cities and insurers assess pollution's true health and financial costs.EN

A Swedish study tracking customer satisfaction across two decades finds that service quality—not product quality—is now the strongest predictor of customer loyalty. The shift has persisted even through recent crises, suggesting companies must prioritize customer experience over product features to maintain competitive advantage.EN