1B iterations was too slow (30s+). Benchmarked on Apple Silicon: - 150M iterations ≈ 6.3s - Falls within the 5-10s target range