Increasing intelligence in AI agents can worsen collective outcomes | ScienceToStartup | ScienceToStartup