I don't, he underdelivered on GPT-5 just like Zuck not a long while ago with Llama 4. Corporations should seriously think about applying novel architectures throughout the next decade in a commercial manner. I'm still praying on infinite context window with full (more or less) accuracy.