Weekly digest for fast-moving tech

NVIDIA’s new agent-safety platform combines open-source runtime controls with independent hardware monitoring to keep autonomous AI systems inside defined security boundaries.

The completed UGV combines modular payloads, damage-tolerant mobility and an aerial-drone launcher as First Hydrogen moves its robotics program into physical hardware.

Researchers reprogrammed E. coli to grow on carbon dioxide and methanol while producing useful chemicals, pointing toward more flexible one-carbon biomanufacturing.

A machine-learning pipeline could reduce hours of manual football-impact video review while keeping researchers focused on sensor events most likely to be genuine.

A physics-informed machine-learning framework predicts how mutations alter protein–RNA interactions, potentially giving researchers a sharper tool for studying molecular variation.

RNA sequencing reveals gene-expression differences associated with engineered malaria mosquitoes, adding another layer to the laboratory assessment of gene-drive systems.

Taste Profile lets Premium users tell Spotify what they want more, or less, of instead of leaving recommendations entirely to inferred listening behavior.

A new architecture aims to give AI persistent memory across devices while keeping the encryption keys and control of personal data with users.

The kine algorithm uses neural fields to reconstruct a continuous video from 116 VLBI observations spanning 27 years, achieving four times the resolution of any single image.

Stanford physicists observed a phonon jumping between energy states in a mechanical resonator, proving sound can behave quantum mechanically and opening a path to phonon-based quantum computing.