The Gaia space telescope works like a sieve, sifting through myriad celestial objects. It gathers photometric data, measuring brightness across different wavelength bands, and spectroscopic data, which splits light into a rainbow spectrum. A neural network trained on this data identifies characteristic patterns: some correspond to stars, others to galaxies. For sufficiently bright sources, this classification achieves over 96% accuracy.

However, faint and distant objects, like the finest grains of sand, slip through the sieve's mesh. That's where the infrared telescope CatWISE2020 comes in: it detects thermal radiation, which can penetrate clouds of cosmic dust. If Gaia is a sieve with relatively large holes, CatWISE2020 acts as a sieve with much finer mesh, catching nearly a third more hidden quasars and galaxies. As a result, the Gaia DR4 catalog has gained three million quasar candidates and two million galaxies, although for the faintest sources there remains a risk of false positives — contaminants mistakenly identified as galaxies.

This grand cosmic census, a tally of celestial bodies, allows astronomers to trace how the universe evolves in its most remote corners. The sieve analogy, of course, simplifies reality: in practice, it involves complex processing of multidimensional data, where each "sighting" is a combination of photometric and spectroscopic features recognized by machine learning.