In this talk I will summarize recent progress and open mathematical problems in understanding black hole mergers quasi-locally, i.e. by using marginally trapped surfaces instead of event horizons.