Core War is a programming game where assembly-like programs, known as warriors, compete for control of a virtual computer. The game operates with a specialized assembly language called Redcode, and warriors are designed to crash their opponents while maintaining their own functionality. The recent work using large language models (LLMs) explores an adversarial evolutionary process called Digital Red Queen (DRQ), which evolves these warriors through a continuous cycle of competition. Each round introduces a new warrior that must adapt to defeat prior versions, resulting in increasingly sophisticated strategies like self-replication and multithreading. The DRQ algorithm mimics biological evolution by pushing warriors to constantly adapt to their environment, which includes all previous iterations. Over time, this leads to the emergence of general strategies that perform well against a range of opponents. Surprisingly, independent runs of DRQ, which start with different initial warriors, tend to converge on similar high-performing behaviors. This convergence occurs at the functional level rather than in the source code, highlighting a form of convergent evolution in programming strategies. The chaotic nature of Core War, where code and data coexist, enables dynamic self-modification. This allows warriors to adapt on the fly, creating a volatile environment where survival depends on rapid evolution. The research suggests that such simulations can provide insights into how AI systems might evolve in real-world adversarial situations, like cybersecurity. The findings also emphasize the potential for LLMs to drive program evolution in a controlled setting, offering a new avenue for studying AI competition.