Just two months into his role as Apple CEO, John Ternus stressed to the software team that Apple has fallen far behind competitors in AI and must accelerate its AI roadmap, moving beyond merely patching Siri. The China-region iOS 27.3 beta updated the Siri interface, but the next-generation AI-powered Siri has yet to launch; the actual experience in regions where related features are already available is underwhelming, offering only screen summaries and single-step commands—far from the level of an agent capable of handling tasks autonomously. To bring AI to China, Apple spent two years selecting Alibaba's Qwen as its cloud model partner, but the actual experience feels stiff, more like a system add-on, still stuck in the old logic of single request and single response, lagging behind the domestic Android camp's long-chain AI operation capabilities. On the earnings call, Ternus publicly affirmed the existing Siri AI, emphasizing the prudent principle of privacy first, but privately he is highly dissatisfied with the software team's methodical pace. Apple once demonstrated a functional prototype close to an agent form, but it was put on hold due to architecture, security, and other issues; it has now adjusted the iOS 28 development model, launched a new framework, and opened related interfaces, attempting to drive exploration of agent architecture. However, the complete personal long-term memory required by an agent fundamentally conflicts with Apple's long-held on-device no-retention privacy principle. Apple is at a critical juncture of leaping from a single-interaction assistant to a fully autonomous agent, and is forced to step out of the comfort zone it has guarded for years and accelerate its transformation.