It is likely that the talk was about multiboundary wormhole/black hole geometries in (2+1)–dimensional gravity.
Remember that BTZ black hole could be seen as a factor space of $\text{AdS}_3$ spacetime by a discrete group of isometries generated by a single element. Multi-black hole geometries are factors $\text{AdS}_3/Γ$ by more complex discrete subgroups $Γ$ of isometry group.
It is possible to think about such geometries in terms of initial data on a slice $t=0$, which would be a fundamental domain of hyperbolic plane $H^2$ under the action of $Γ$:
An more detailed introduction could be found in this blogpost by Shaun Maguire
While such solutions are known for more than 20 years [$1$], recently they attracted attention as an important model to understand entanglement entropy within holography [$2$] and as a realization of $\text{ER} = \text{EPR}$ conjecture [$3$].
Note, that I would argue against statements that second (and third etc.) black hole is inside the event horizon of the first black hole, “behind” would probably be more accurate, since the “world” with second black hole is “hidden” by the horizon of the first (just like the “left” asymptotic region of Kruskal–Szekeres extended solution is hidden from the “right” asymptotic region), but the interior proper of the first black hole does not contain the second.
References
Brill, Dieter R. Multi-black-hole geometries in (2+ 1)-dimensional gravity. Physical Review D 53.8 (1996): R4133, doi:10.1103/PhysRevD.53.R4133, arXiv:gr-qc/9511022.
Maxfield, H. (2015). Entanglement entropy in three dimensional gravity. Journal of High Energy Physics, 2015(4), 31, doi:, arXiv:1412.0687.
J. Maldacena and L. Susskind, Cool horizons for entangled black holes, Fortsch. Phys.61(2013)781–811, doi:10.1002/prop.201300020, arXiv:1306.0533.