AI reconstructs high-energy plasma jets from a distant blazar

A team of Caltech astronomers has synthesized 27 years of radio telescope data into a high-resolution video of a supermassive black hole. By applying a custom neural network to archive imagery of the blazar 3C 345, researchers mapped the trajectory of plasma jets erupting 5.5 billion light-years from Earth.

Editorial·September 24, 2026·reads: 239

AI reconstructs high-energy plasma jets from a distant blazar

The study, published in Nature, marks a shift in astrophysical imaging. Instead of relying on conventional pixel-based processing, lead researcher Dr. Marianna Foschi and her team deployed a neural-network algorithm named Kine. This software analyzed 116 images captured between 1995 and 2022, filling visual gaps to improve image clarity by fourfold.

This method transforms isolated snapshots into a continuous record of plasma motion. By mapping the blazar’s brightest peaks and darkest valleys, the team achieved a level of resolution typically reserved for next-generation hardware. Dr. Foschi noted that the algorithm extracts these insights from existing data, effectively bypassing the wait for more powerful sensors or lenses.