At the center of a galaxy, where a supermassive black hole spins a disk of scorching gas around itself, giant stars are sometimes born. They are unlike ordinary stars: the disk continuously feeds them with matter, and a star can grow to hundreds of solar masses, staying bright as long as the gas flow lasts. Such a star is like a sponge soaking up water: it lives while being filled. If a stellar-mass black hole ends up in this sponge, something unusual begins.

The black hole, once inside the star, does not fall straight to the center. The gas around it slows its motion just as viscous honey slows a falling lead shot. This gas drag makes the black hole spiral inward, gradually approaching the star's core. The journey from the surface to the center takes about a tenth of a year — a blink of an eye by cosmic standards. If a second black hole enters the star, they meet in the core and form a tight pair.

The pair of black holes inside the star behaves like two drops of mercury dropped into oil. The surrounding gas does not just slow them: it squeezes energy out of the orbital motion, forcing the black holes closer together. The merger time shrinks from millions of years to one hour. Gravitational waves from such a merger would be noticeable to LIGO–Virgo–KAGRA detectors, but the signal has a peculiarity: at frequencies that LISA can hear, it is suppressed, and in the decihertz range, it is phase-shifted. The future DECIGO detector could pick out this signature.

Each giant star undergoes only one or two such mergers. After that, the gas inside it heats up or the star is destroyed, so hierarchical growth of black holes to intermediate masses is limited. This explains why, in the population of mergers detected by LIGO–Virgo–KAGRA, there is an excess of events with positive effective spin: black holes that merged inside a star inherit its rotation.

Let's return to the sponge soaking up gas. A giant star in an active galactic nucleus disk is not just a passive reservoir. It becomes a natural merger accelerator, where gas plays the role of an invisible hand bringing black holes together in hours. This mechanism may have already left a mark in the data of gravitational-wave observatories, but to see it, we need next-generation detectors sensitive to decihertz frequencies. Then we could peer inside these stellar cocoons and see how pairs of black holes are born.