Great to see this trend of thinking about AI issues (capabilities, alignment, safety) in terms of an ecosystem of AIs rather than single models in isolation.