2026年9月10日

Honor Engineers on the Doubao Smartphone: Pioneering New Interaction Models in the Age of AI

Fast Technology reported on December 7 that ByteDance has officially released a technical preview of...

Fast Technology reported on December 7 that ByteDance has officially released a technical preview of its Doubao AI Mobile Assistant. The assistant has been pre-installed on the Nubia M153 engineering prototype, a device co-developed with ZTE, and is currently being offered in limited quantities.

Commenting on the emergence of the Doubao phone, a performance optimization R&D engineer from Honor, known as “MagicOS Performance – A Yong,” stated that the Doubao phone represents a bold exploration of new interaction models in the AI era. He noted that it attempts to deeply integrate AI capabilities into a smartphone brand’s closed-loop hardware ecosystem, signaling a meaningful shift in how users may engage with mobile devices in the future.

However, he also pointed out that many applications are already imposing restrictions on its usage. From the current outcomes, he believes that it will be far from easy for internet companies to build strong competitive moats by leveraging existing ecosystems to accelerate growth.

In his view, the future of AI smartphones will inevitably see the emergence of more “experience closed loops.” Building such an AI-driven experience ecosystem is not something that can be achieved overnight—it requires long-term strategic patience and continuous refinement.

Meanwhile, the Doubao mobile assistant team recently announced that further usage restrictions will be implemented. These include limiting score farming and reward farming behaviors, restricting certain gaming scenarios, and temporarily suspending the assistant’s ability to operate banking and online payment applications.

The team also emphasized that they are actively engaging in in-depth discussions with various app developers, with the goal of establishing clearer and more predictable rules. They hope to avoid a one-size-fits-all approach that could undermine users’ legitimate rights to use AI in reasonable and practical ways.

接著讀