Archived Intelligence

27 March 2026

Lead StoryEnvironmental Science
Decoding the African Humid Period to Predict Future Droughts

Decoding the African Humid Period to Predict Future Droughts

Researchers analysed a 10,250-year sediment record from Lake Yoa in Chad to map sudden droughts during the ancient greening of the Sahara. The findings suggest that rapid freshwater inputs in the North Atlantic triggered decadal droughts in Africa, underlining the critical need for improved future climate forecasting.

Full Report

Image for The Factory Cleanup Crew: How a New KRAS G12D Targeted Therapy Destroys Cancer Proteins
Medicine & Health

The Factory Cleanup Crew: How a New KRAS G12D Targeted Therapy Destroys Cancer Proteins

A phase 1 trial of setidegrasib, a first-in-class protein degrader, shows promising antitumour activity against lung and pancreatic cancers. By tagging the faulty protein for destruction, this approach could offer a clever new option for hard-to-treat tumours.

Image for How Earth's Magnetic Shield Blocks Galactic Cosmic Rays Near the Moon
Physics & Astronomy

How Earth's Magnetic Shield Blocks Galactic Cosmic Rays Near the Moon

Data from the Chang'E-4 lunar lander reveals that Earth's magnetic field creates a cavity of reduced radiation near the Moon. This natural shield could protect astronauts on future lunar missions by providing a safer zone from deep-space radiation.

Image for The Next Decade of Chemical Synthesis: Catalytic Hydrogenation of Nitroarenes Using Manganese
Chemistry & Material Science

The Next Decade of Chemical Synthesis: Catalytic Hydrogenation of Nitroarenes Using Manganese

Researchers have successfully developed a manganese-based single-atom catalyst for the selective synthesis of amines. This approach replaces rare, expensive metals with abundant materials, potentially lowering the cost of global chemical manufacturing.

Image for The Violent Winds Tearing Through Starburst Galaxies
Physics & Astronomy

The Violent Winds Tearing Through Starburst Galaxies

Astronomers have measured the extreme velocity and temperature of gas in the core of M82. The findings suggest that thermal pressure from supernovae alone is powerful enough to drive massive galactic winds, solving a long-standing cosmic mystery.

Image for Renewable energy optimization: How AI could stabilise the future grid
Computer Science & AI

Renewable energy optimization: How AI could stabilise the future grid

Researchers have developed a hybrid machine learning model that predicts and manages the unpredictable nature of wind and solar power. By combining neural networks with nature-inspired algorithms, this system processes grid variables significantly faster than traditional methods.

Image for How Machine Learning for Perovskites Could Fix Solar Energy's Toxic Bottleneck
Chemistry & Material Science

How Machine Learning for Perovskites Could Fix Solar Energy's Toxic Bottleneck

A new review outlines how algorithms can predict the properties of perovskite-inspired materials. This data-driven approach could vastly accelerate the discovery of stable, non-toxic solar panels.

Image for The Future of Cellular Medicine: A New Standard for hiPSC Genetic Stability
Genetics & Molecular Biology

The Future of Cellular Medicine: A New Standard for hiPSC Genetic Stability

Current methods for checking stem cell safety often miss hidden genetic errors. A new multi-tool approach provides a comprehensive check, ensuring laboratory-grown cells remain safe for patient use.

Image for The Quiet Brilliance of On-chip frequency combs
Physics & Astronomy

The Quiet Brilliance of On-chip frequency combs

Researchers have successfully integrated a semiconductor laser with a nanophotonic circuit to create stable, microscopic pulses of light. This advance bypasses the need for fragile resonators, offering a robust path toward portable precision metrology.

Image for The Electricity Metre of AI: The Hidden Challenge of LLM Safety Auditing
Computer Science & AI

The Electricity Metre of AI: The Hidden Challenge of LLM Safety Auditing

Researchers tried to detect when an AI refuses a dangerous prompt just by monitoring computer hardware signals. They discovered that without looking inside the software, a computer's hardware cannot tell the difference between safe and unsafe tasks.