According to Ming-Chi Kuo, OpenAI is accelerating plans for its own hardware with a mass production target in H1 2027. Instead of an app-centric OS, the entire device will run natively on persistent AI agents executing multi-step tasks in the background. Leaked Specs: Custom MediaTek Dimensity 9600 chip (TSMC 2nm N2P process) Heterogeneous dual-NPU setup (processing vision + language simultaneously) LPDDR6 RAM + UFS 5.0 to handle local model inference bottlenecks Kuo projects around 30 million s...