Study Finds AI May Be Accelerating Its Own Development

Artificial intelligence may already be accelerating the development of more capable AI systems, according to an independent assessment by the nonprofit research organization METR. Researchers examined confidential performance data, interviews and internal research from Anthropic during an 11-week investigation. They estimated that Anthropic’s models have been improving on well-defined benchmark tasks approximately 1.9 times faster since October 2025 than during the preceding period. METR suggested the acceleration reflects AI’s growing contribution to coding, experimentation and model development rather than faster increases in training computing power. However, the estimate carries considerable uncertainty, with plausible acceleration ranging from 1.5 to 2.5 times the earlier rate. Reported November 5, the findings raise questions about whether AI safety evaluations can keep pace with increasingly rapid technological progress. (Reuters)

Scientists Discover the Brain Imports Protective Proteins From Other Organs

Researchers at the Gladstone Institutes have created the first atlas of blood proteins that cross the blood-brain barrier, revealing an unexpectedly sophisticated delivery system that may play a role in Alzheimer’s disease. Using mice, scientists tracked 861 circulating proteins and found that 92 entered the brain through selective transport mechanisms. Many originated in distant organs, including the liver. One protein, APOA1, helped maintain connections between neurons by supporting astrocytes. In mouse models of Alzheimer’s, the barrier stopped importing protective proteins before it became physically leaky, suggesting an early failure in brain maintenance. Increasing liver production of APOA1 reduced Alzheimer’s-related abnormalities in the animals. Published in Cell and reported November 11, the findings could inspire treatments that restore protective protein delivery. Human effectiveness remains untested. (Medical Xpress)

AI Climate Models Could Become Faster, Cheaper and More Reliable

Scientists at MIT have developed a new training method that could make artificial intelligence climate models substantially faster to produce and more dependable when predicting unfamiliar future scenarios. Conventional Earth-system models can require months of supercomputer processing, while simplified AI emulators generate projections in seconds. However, existing emulators often struggle when emissions or environmental conditions differ from their training examples. The MIT team addressed this problem by creating optimized synthetic training scenarios containing rapid variations in greenhouse gas emissions. Their approach required only one expensive Earth-system simulation rather than the six to eight typically used for training. The resulting emulator successfully predicted previously unseen scenarios, including unusual combinations of pollution-control policies. Published in Proceedings of the National Academy of Sciences and reported November 12, the advance could improve climate-policy planning. (MIT News)

European Spacecraft Begins Close Inspection of NASA’s Asteroid Impact Site

The European Space Agency’s Hera spacecraft has begun examining the asteroid system struck by NASA’s DART mission in 2022, marking an important new stage in planetary-defense research. Hera is observing the larger asteroid Didymos and its smaller companion Dimorphos from approximately 25 to 30 kilometers away. NASA deliberately crashed DART into Dimorphos to test whether a spacecraft could redirect a potentially dangerous asteroid. The collision shortened Dimorphos’s orbit and altered its path around the Sun, with ejected debris providing additional momentum. Hera carries 25 scientific instruments and two small CubeSats designed to investigate the impact’s aftermath. The spacecraft is expected to approach within approximately 20 kilometers in December. Announced November 11, the observations should help scientists understand how different asteroid structures respond to impacts. (Associated Press / Phys.org)

Engineers Discover That Disorder Can Make Materials Stronger

Engineers at MIT and Northeastern University have developed a new class of metamaterials that become more resilient when their internal structures are deliberately irregular. Conventional metamaterials typically contain repeating geometric units arranged in orderly patterns. The researchers instead designed networks of interconnected beams with different sizes and thicknesses, inspired by leaf veins, dragonfly wings and complex natural networks. Using algorithms and 3D printing, they produced polymer structures that distributed damage more effectively and resisted compression and sudden impacts better than conventional ordered designs. The team also demonstrated a protective structure capable of cushioning a fragile object within its interior. Published November 11 in Nature, the findings suggest that carefully engineered disorder could improve protective helmets, packaging, armor and materials designed to control sound or vibration. (MIT News)

New MRI Technology Could Spare Cancer Patients Aggressive Radiation Treatment

Researchers at London’s Institute of Cancer Research are developing an MRI technique that could help doctors reduce unnecessary radiation exposure in patients with HPV-positive throat cancer. Called HYPOXiMAP, the technology creates color-coded images showing oxygen levels inside tumors. Oxygen-starved regions, displayed in blue, tend to resist radiotherapy, while well-oxygenated areas, shown in yellow, are generally more responsive. Researchers have demonstrated that the technique can distinguish these regions and monitor changes during treatment. A new four-year program will incorporate artificial intelligence and test the method across multiple hospitals. Reported November 10, the advance could eventually help physicians tailor radiation doses while reducing swallowing difficulties, dry mouth and other lasting complications. However, clinical trials must still establish whether MRI-guided dose reductions preserve cancer cure rates. (Medical Xpress)

Scientists Identify Hidden Heart Attack Risks Beyond Bad Cholesterol

Lowering LDL cholesterol substantially reduces cardiovascular risk, but scientists have identified additional mechanisms that may explain why heart attacks still occur in patients whose cholesterol levels are well controlled. Researchers at Sweden’s Chalmers University of Technology analyzed nearly 3,000 blood proteins alongside genetic information and long-term health records. They discovered that triglyceride-rich lipoproteins and lipoprotein(a) share protein signatures associated with inflammation and changes to blood-vessel walls. These mechanisms differ from the cholesterol-deposition pathways primarily associated with LDL particles. Published in the European Heart Journal and reported November 10, the findings suggest that certain fat-carrying particles may damage arteries through overlapping inflammatory processes. The research could inform more personalized cardiovascular prevention, although no new treatment or diagnostic test was evaluated. (ScienceDaily)

Ancient Armored Amphibian Reveals a Lost Experiment in Ocean Evolution

Paleontologists have discovered a remarkable amphibian that apparently lived entirely in the ocean approximately 237 million years ago, challenging assumptions about the ecological limitations of amphibians. Named Gerrothorax maritimus, the animal measured nearly three feet long and possessed a broad, flattened body covered in protective bony armor. Two exceptionally preserved skeletons recovered from China’s Guizhou Province were found in deep-marine sediments. Chemical analysis of their strontium isotopes, combined with the preservation of delicate bones, supports the interpretation that these animals lived and died at sea rather than being transported from freshwater environments. Published in Nature Ecology & Evolution and reported November 10, the discovery suggests some ancient amphibians evolved adaptations for saltwater survival, possibly using armored skin and gills to regulate water and salt balance. (ScienceDaily)

NASA Discovers Hidden Companion Behind Star’s Repeated Explosions

Astronomers using NASA’s Chandra X-ray Observatory have identified the likely cause of mysterious eruptions from a massive star located approximately 30 million light-years away. The star, AT 2016blu, has produced at least 27 dramatic brightness increases since 2012, recurring roughly every 113 days. Although the outbursts resemble supernova explosions, the star survives each event. New X-ray observations suggest that an unseen companion, either a neutron star or stellar-mass black hole, periodically captures material from the unstable star as the objects approach one another. The infalling material becomes extremely hot and produces intense X-rays. Published in The Astrophysical Journal and announced November 9, the findings provide the first X-ray detection of this class of erupting stars and could help explain other apparent stellar explosions. (NASA)

Astronomers Discover Two Rings of Planetary Wreckage Around Dead Star

Astronomers have discovered an unusual double-ring structure surrounding a white dwarf, offering a glimpse of what can happen to planetary systems after their stars die. The object, J1328−1450, lies approximately 1,600 light-years away in the constellation Virgo. Observations using the European Southern Observatory’s Very Large Telescope revealed two distinct rings of gaseous debris, unlike the single discs previously identified around other white dwarfs. Researchers suspect that surviving rocky bodies, possibly including one approximately Pluto-sized object, helped shape the rings after the original star expanded into a red giant. Differences in magnesium abundance suggest the rings may contain material from separate disrupted objects. Published in Nature Astronomy and reported November 7, the findings suggest planetary systems can retain unexpectedly complex structures long after their stars exhaust their fuel. (Reuters)