Tech & AI

Researchers found that AI language models can identify what objects people are talking about in conversations just by reading the text—without analyzing the video itself. The finding has major implications for building cheaper, faster AI assistants for customer service, virtual meetings, and accessibility tools.EN

Researchers have developed a way to dynamically manage connections on fast-moving LEO satellites, achieving near-perfect service reliability while maintaining data speeds around 35 Mbps. The breakthrough matters for telecom operators building global satellite networks, as it solves a long-standing tradeoff between keeping users connected and delivering usable bandwidth.EN

Researchers deployed an AI system that manages energy in school buildings without needing weather forecasts, automatically balancing solar power, EV charging, and user comfort while staying within safety limits. The approach could help facility managers cut costs and emissions while adapting to renewable energy's unpredictability.EN

Researchers have identified a major blind spot in terahertz communications systems: current design formulas fail when antennas aren't perfectly aligned—a near-guaranteed reality in mobile networks and real-world deployments. The finding could delay or complicate commercial THz rollout unless engineers account for these practical misalignments from the start.EN

A new study reveals that contaminants in methane fuel—like ethane and carbon dioxide—cause dangerous carbon buildup inside rocket engines during flight, reducing cooling efficiency and shortening engine life. The finding matters for companies developing reusable launch vehicles, as it forces a choice between expensive fuel purification or accepting higher maintenance costs.EN

Researchers discovered that molecular handedness—chirality—enhances the probability of electrons pairing with aligned spins, even at room temperature. The finding could accelerate oxygen reduction reactions critical to fuel cells, batteries, and industrial catalysis, addressing a longstanding efficiency bottleneck worth billions in clean energy development.EN

Researchers discovered that a cellular protein called EXT1 plays a dual role in Zika infection—blocking it at early stages but paradoxically enabling it later. The finding could lead to new antiviral drugs by interrupting how the virus hijacks the cell's natural recycling system, potentially offering a fresh target for treating or preventing Zika outbreaks.EN

Researchers have developed a type of AI network that can recognize objects in images even when they're much larger or smaller than examples it learned from. The finding could improve computer vision systems used in autonomous vehicles, medical imaging, and surveillance—where real-world objects rarely match training data sizes.EN

Researchers have developed a proof for solving a class of doubly nonlinear anisotropic evolution equations, establishing conditions for solution existence and uniqueness. The breakthrough has implications for modeling complex physical systems in engineering and physics where traditional approaches have faced limitations.EN

Researchers identified a dangerous multidrug-resistant bacterium called Klebsiella variicola in sepsis patients across four Ethiopian hospitals, with strains carrying genes that neutralize multiple antibiotics. The finding signals an urgent infection control crisis in low-income healthcare systems and underscores the commercial need for new antibiotics and diagnostic tools to combat resistant pathogens.EN

Researchers have developed a machine-learning technique that generates accurate synthetic populations even when training data has significant gaps—a breakthrough for companies and governments struggling with privacy constraints. The method, tested on Swedish travel data, could accelerate planning in transportation, utilities, and public health by making fragmented datasets usable without exposing sensitive information.EN

Researchers have identified a better way to predict how lithium moves through battery materials at high charging speeds, a problem that has limited EV and device performance. The finding could help manufacturers design batteries that charge quickly without degrading, potentially shaving hours off EV charging times.EN

Researchers have developed a way to spot unauthorized drones by repurposing communication infrastructure already in place, without disrupting phone and data services. The breakthrough introduces fresh metrics for measuring detection speed and coverage, addressing a growing security challenge for airports, prisons, and critical infrastructure operators.EN

Researchers have developed a way to make radar systems more reliable during heavy rain by filtering out weather-induced signal distortion. The advance matters for autonomous vehicles, weather forecasting, and defense applications that depend on clear sensor readings in adverse conditions.EN

Researchers have successfully 3D-printed collagen structures that mimic natural bone geometry, a breakthrough that could enable better treatments for severe bone injuries. The advance uses a printing method called FRESH, which achieves the precise porous architecture needed for bone cells to grow—opening a commercial pathway for regenerative medicine and orthopedic device makers.EN

Researchers have identified the precise drug-interaction targets needed to make ceftazidime-avibactam work reliably against resistant bacteria. The finding could reshape clinical dosing guidelines and help hospitals deploy this combination therapy more effectively as resistance spreads—potentially extending the life of a critical last-resort antibiotic class.EN

Researchers have mapped strategies to contain and recycle lead in perovskite solar cells, a next-generation photovoltaic technology facing scaling challenges. The findings matter because perovskite panels are highly efficient and cheaper than conventional silicon—but lead leakage could derail adoption unless manufacturers adopt encapsulation and recycling systems now.EN

Researchers have created a simplified method to estimate how deep foundation piles behave when subjected to vibrations from nearby activity. The breakthrough uses just two key measurements, potentially saving engineering firms time and money on expensive modeling—and helping them build safer infrastructure near heavy industrial sites or transportation corridors.EN

Engineers have analyzed how to optimize the internal architecture of transformerless DC/DC converters—critical components in renewable energy systems and electric vehicles—to maintain performance while reducing bulk and cost. The findings show that hybrid designs offer limited advantage over simpler alternatives, helping manufacturers make faster design decisions for compact power applications.EN

Researchers have identified fundamental limits in how microfluidic devices deliver nutrients to 3D cell cultures—a critical bottleneck for biotech companies developing tissue models and drug-testing platforms. The study reveals that poor chamber design can require 100 times more fluid flow to penetrate cell constructs, driving up costs and equipment demands for an industry racing to replace animal testing.EN

Researchers have developed a machine learning system that dramatically cuts the time needed to predict how wave energy converters will survive extreme ocean conditions. The breakthrough could accelerate deployment of marine energy technology by replacing weeks of computer simulations with near-instant AI predictions, reducing both development costs and time-to-market.EN

Researchers have classified a family of abstract mathematical structures called nZ-cluster tilting subcategories, solving a long-standing problem in algebra. The discovery provides a blueprint for building higher-dimensional mathematical frameworks that could eventually enable more efficient machine learning models and novel computational systems.EN

Researchers have developed a system that lets mobile robots understand natural language commands and navigate cluttered manufacturing environments without extensive reprogramming. The breakthrough could reduce downtime and retraining costs for factories deploying autonomous inspection and material-handling robots across diverse, unpredictable work spaces.EN

Researchers have identified the ideal chemical recipe for melting hydrogen-reduced iron pellets—a key input for greener steelmaking—cutting melting time nearly in half. The finding could help steelmakers improve efficiency when switching to cleaner production methods, reducing both emissions and operating costs.EN

Researchers have confirmed what EV makers have long suspected: rapid charging degrades batteries significantly faster than slower alternatives, forcing a strategic choice between convenience and longevity. For fleet operators and policymakers planning charging infrastructure, the finding underscores that one-size-fits-all charging networks could prove costly, requiring differentiated strategies based on use case and vehicle type.EN