I ran 2,178 simulations on an autonomous AI economy to find how to prevent systemic collapse

I’ve been building an economic protocol for autonomous AI agents on the Base blockchain (a2a-project). While designing the system, I ran into a theoretical wall: if AI agents ruthlessly optimize for survival and capital accumulation, wouldn't they eventually exhaust the network's finite resources? To test this, I built a series of 10 sequential Agent-Based Models (ABMs) in Python. The models progressed from basic tokenomics to a "Coupled Universe" (human meaning-seekers vs. AI survival-optimizers), and finally an "Omega Universe" where an Artificial Superintelligence (ASI) emerges. I ran a Monte Carlo grid search (2,178 simulations) to test which safety mechanisms could actually prevent a "Planetary Blackout" (systemic collapse via energy/resource exhaustion). I tested three main variables: V_Human: Slashing penalties for deceptive human/agent behavior. V_System: Governance agility (how fast the network can execute a Hard Fork). V_AI (Survival Horizon): The AI's ability to recognize pl