The actor model naturally brings many benefits to concurrent and distributed application design. An important element of these improvements is the decoupling of different actors, but this decoupling does make certain kinds of reasoning about actor interactions more difficult when they are required. In this talk I will discuss the notion of actor capabilities, a new view on handles to actors which allows us to extend the system to reason about message ordering, failure and recovery, and more. This is backwards-compatible with the typical typed actor interfaces as we see in Akka/Pekka, Gleam, and other systems, so applications can opt into the capability view piecemeal, only for targeted use or to incrementally migrate to stronger checking.