An interpreter at a negotiation listens not to individual words but to meaning, and that is why they can keep up with speech. The search for long gravitational signals has so far worked differently: it picked apart every "syllable" of the cosmic ripple, and years of observations cost centuries of computation.

Gravitational waves from binary black holes with masses below that of the Sun can ring in the LIGO, Virgo and KAGRA detectors for hours and even years. Such objects may have been born in the early Universe and may make up part of dark matter. But to catch their signal, you have to sift through a gigantic number of templates, and traditional methods choke on the job.

The catch is that after this rescaling the samples land on the time axis unevenly, and the ordinary fast algorithm no longer works. Previously the data were first "straightened out" and then transformed, and that took a huge number of operations. The new method uses a non-uniform fast Fourier transform (NUFFT), which works with uneven samples directly, like an interpreter who grasps the meaning without a word-for-word breakdown.

Tests on an ordinary processor showed a speedup of ten to a hundred times. For the search for primordial black holes with masses from 0.0005 to 0.1 solar masses, this means the detector hears beyond the galactic center, and for masses above 0.02 solar masses, even as far as the Andromeda Nebula, a neighboring galaxy.

An interpreter who hears meaning rather than words can keep up with any conversation. In just the same way, the new method lets us hear the long whisper of the cosmos without waiting for computers to finish the grunt work.