Or OpenAI is? After using Gemini 2.5, I did not feel "AI is slowing down". It's just this model isn't SOTA.