Anthropic Needs to have the most intelligent and scary agents, so if OpenAI does something bad they need to prove their models can do even worse. Without that the whole valuation collapses. There’ll be more “our model outhacks others” for the next hype cycle.