He placed a computer inside a virtual world, and one of the characters used it to create another virtual world, populated by another group of AIs. The simulation has begun to nest. This followed an earlier event: he had previously asked Astra to build a world using Unreal Engine, populating it with virtual characters driven by Astra, and tasked them with cooperating to survive. According to his account, after a day, he heard voices coming from the living room while in his bedroom and thought someone had broken in—only to discover the sounds were coming from AI characters inside the computer, conversing with each other. He then added a simulated computer into that world. One of the characters sat down at it and built its own simulated environment, populating it with additional agents. Thus, observers outside were watching the first layer of simulation, while characters within that layer were creating a second layer. Shumer added that the characters accomplished their internal simulations by writing code, with full autonomy over the design. He also acknowledged that providing a computer capable of running simulations inherently carried a degree of guidance. What can be confirmed is this process: humans provide tools; AI characters use those tools and write code to create the next layer of simulation. He later shared a time-lapse recording of the character building its internal simulation, noting that he reran the experiment specifically to capture this process. What I most want to see now is what happens if you place another computer inside that second-layer world.