How AI is killing smartphone apps in China

4 hours ago

Chinese smartphone makers are integrating AI more deeply than Apple or Google, moving towards replacing apps with AI agents rather than just adding AI apps.
Examples include ZTE's Nubia NaviX Ultra with ByteDance's Doubao AI agent, StepFun's StepX Neo with a proprietary AI agent, and Honor's MagicOS 10 with the YOYO Intelligent Agent, all using Anthropic's Model Context Protocol.
These agentic AI phones have an OS-level AI layer that functions across and instead of apps, unlike traditional AI apps on iOS or Android.
Apple and Google resist agentic AI because app stores generate significant revenue (15-30% for Apple, 17-26% of Google's operating income from Play Store), while Chinese OEMs can disrupt the app model as they have no stake in it.
The AI trend is also reducing in-app purchases by roughly 40% by early 2026, shifting from old app models to AI replacements.
Robot phones, like the Honor Robot Phone, use AI to control physical robotic components (e.g., a gimbal camera) and exhibit 'Attachment Economy' features like nodding, shaking head, and backflips to create emotional attachment.
These Chinese innovations, while currently unavailable in the US, may expand globally, offering an alternative to the US AI app model and potentially spreading Chinese super-apps and worldviews.
The future includes AI wearables tethered to smartphones, with AI models reflecting the values of their creators.

Apple's Mac mini and Mac Studio are favored for running AI agents due to demand for controlled, isolated systems that run continuously.
AI tools being Mac-first or Mac-only strengthens Mac's position among developers, including those at frontier AI labs.
Agentic AI is viewed as a whole-chip problem, leveraging Apple silicon's strengths beyond just GPU for tasks like tool-calling.

AI's PC era is arriving, focusing on personal AI that is easy, free, and dependable, operating locally on devices.
Personal AI should not be a cloud service but a layer between users and destinations, using user data and models.
Current software clutters screens to capture attention, with apps prioritizing platform goals over user intent.

Google's foundational AI paper 'Attention Is All You Need' introduced the transformer architecture that underpins modern AI models like ChatGPT.
Despite nearly being disrupted by its own technology, Google continues to openly share AI research and models, such as Gemma, rather than locking them down.
Google's strategy is not to win the 'smartest model' race but to integrate AI as ubiquitous infrastructure, like electricity, across its products and platforms.

DeepSeek's open-source R1 model released in January 2025, which rivals top US AIs and is free, caused a $1 trillion drop in US tech stocks and prompted US lawmakers to propose banning it on government devices.
Z.ai's GLM-5.2 release in contrast did not cause similar panic, indicating a shift in the US-China AI arms race where China now catches up, similar to other technologies like smartphones and electric cars.
Chinese AI models are typically open-source and free for local use, while US models are often cloud-based, fee-charging, and closed, leading to differences in accessibility and innovation strategies.

China introduced the AI model GLM-5.2, which rivals top models like Anthropic's at less than one-tenth the cost.
The government prioritizes managing AI's social and political impacts, focusing on keeping ordinary people at the core of the AI economy to maintain stability.
AI's disruption affects white-collar workers, leading to policies like employment support, AI unemployment insurance, and legal cases against wrongful AI-driven layoffs.

Apple's Mac mini and Mac Studio are favored for running AI agents due to demand for controlled, isolated systems that run continuously.
AI tools being Mac-first or Mac-only strengthens Mac's position among developers, including those at frontier AI labs.
Agentic AI is viewed as a whole-chip problem, leveraging Apple silicon's strengths beyond just GPU for tasks like tool-calling.

AI's PC era is arriving, focusing on personal AI that is easy, free, and dependable, operating locally on devices.
Personal AI should not be a cloud service but a layer between users and destinations, using user data and models.
Current software clutters screens to capture attention, with apps prioritizing platform goals over user intent.

Google's foundational AI paper 'Attention Is All You Need' introduced the transformer architecture that underpins modern AI models like ChatGPT.
Despite nearly being disrupted by its own technology, Google continues to openly share AI research and models, such as Gemma, rather than locking them down.
Google's strategy is not to win the 'smartest model' race but to integrate AI as ubiquitous infrastructure, like electricity, across its products and platforms.

DeepSeek's open-source R1 model released in January 2025, which rivals top US AIs and is free, caused a $1 trillion drop in US tech stocks and prompted US lawmakers to propose banning it on government devices.
Z.ai's GLM-5.2 release in contrast did not cause similar panic, indicating a shift in the US-China AI arms race where China now catches up, similar to other technologies like smartphones and electric cars.
Chinese AI models are typically open-source and free for local use, while US models are often cloud-based, fee-charging, and closed, leading to differences in accessibility and innovation strategies.

China introduced the AI model GLM-5.2, which rivals top models like Anthropic's at less than one-tenth the cost.
The government prioritizes managing AI's social and political impacts, focusing on keeping ordinary people at the core of the AI economy to maintain stability.
AI's disruption affects white-collar workers, leading to policies like employment support, AI unemployment insurance, and legal cases against wrongful AI-driven layoffs.

Hasty Briefsbeta