The immediate predecessor. Concept-injection experiments showing models can, sometimes, accurately report on their own internal states. The global-workspace paper takes this from "can it introspect?" to "what is the structure it introspects on, and is it causally real?"