I built a simulator that runs AI regulations through 10,000 agents and shows you how many comply, relocate, and who evades

SwarmCast is a simulation tool that runs AI regulations through 10,000 heterogeneous agents (companies, startups, regulators, investors) across 15 jurisdictions to model compliance, evasion, and jurisdiction flight patterns. It features epistemic honesty tagging (GROUNDED/DIRECTIONAL/ASSUMED) and adversarial injection to test governance resilience under coordinated narrative pressure. The vectorized engine processes 10K agents in ~3 seconds, with an optional LLM swarm mode using 23 persona agents for deeper reasoning.