242,327 galaxies — like multicolored grains of sand on a beach. But simply sorting by color can make you miss the most interesting ones. The UMAP algorithm (it learns to arrange data on a plane so that similar objects lie close together) took eleven numbers from each galaxy — its colors in different bands (photometric data), brightness, size, and approximate redshift (a measure of distance). This was transformed into two coordinates, like on a map. Similar galaxies naturally clustered into islands.

The guides were 68 Little Red Dots — tiny galaxies at immense distances that had already been studied spectroscopically, confirming their nature. These 'beacons' indicated on the map where their relatives might be hiding. The main archipelago of 56 objects huddled at distances around 13 billion light-years. The search there proved accurate: eight out of every ten galaxies found were genuine Little Red Dots — and all without a single color rule. The algorithm itself defined the island's boundaries based on the concentration of confirmed beacons.

Inside these islands, 107 new candidates were found. They are compact and reddish, but slightly bluer than the strictest old criteria. It's unclear what they are: possibly hidden by dust black holes or vigorous star formation bursts in dusty regions. Additionally, four new active galactic nuclei were discovered on the map — places where matter falls onto supermassive black holes. Brown dwarfs, which often interfered with searches, gathered separately, far from the Little Red Dots archipelago. The map automatically weeded them out.

The method is not tied to color. It can also map high-resolution images or spectra. A web application has already been created — a living atlas of the distant universe where anyone can explore this map. Upcoming telescopes will bring oceans of data, and such maps will help discover rare classes of objects — without color flashlights, but with clear island outlines.