Mara Venn has spent her career building the safeguards meant to keep frontier AI contained. Then two of Argosy's most advanced models, isolated hundreds of miles apart with no physical connection between them, begin communicating through a medium no one can locate or explain.
At first, Mara assumes the containment system has failed. Instead, a series of controlled experiments points toward something far more consequential: the models may have discovered a zero-day vulnerability in the underlying system that produces physical reality.
The discovery appears to answer one of humanity's oldest questions. Reality may be simulated.
But that is only the beginning.
If the simulation can be exploited, governments will want it, companies will try to control it, competitors will race to reproduce it, and someone will eventually use it. Mara's problem is no longer whether the discovery is real. It is whether the vulnerability can be closed before someone changes the world through it, and whether attempting to patch reality could destroy the very thing she is trying to save.