At the end of 2025, the debut of a smartphone sent shockwaves through the entire internet industry. The Doubao phone (nubia M153), jointly launched by ByteDance and ZTE, sold out 30,000 technical preview units within 117 seconds, with resale premiums exceeding RMB 12,900. Behind this explosive market enthusiasm lies not an extreme stacking of hardware specifications, but a fundamental disruption of the underlying logic of the mobile internet. It is no longer a cold carrier for Apps, but a system that embeds the Doubao large model as an AI Agent into the core of the operating system. With system-level INJECT_EVENTS permissions, it tears open the “island economy” defense built by traditional Apps. A battle over control of the user entry point has officially begun.
This transformation delivers a comprehensive and potentially fatal blow to traditional internet giants. For years, super Apps have monopolized user attention through “home-screen dominance,” with downloads, open rates, and dwell time serving as core value metrics. Giants such as Baidu, Alibaba, and Tencent built seemingly unassailable traffic moats on this logic. Within the Doubao phone ecosystem, however, users no longer need to tap App icons. Actions such as “one-click triple engagement on a Bilibili video” or “editing a photo and posting to Moments” can all be completed via the AI assistant, instantly rendering long-revered core metrics obsolete. More devastating still is the collapse of commercial logic. App-based advertising monetization depends on users browsing within applications, yet the Doubao phone’s “cross-app penetration” capability allows users to purchase products without browsing e-commerce homepages, or order food without scrolling through delivery feeds. A sharp decline in ad impressions becomes inevitable. Zhou Hongyi, founder of the 360 Group, has bluntly stated that this model undermines the traditional traffic logic relied upon by major platforms, placing their core KPIs at risk of failure. The dismantling of data barriers makes the shock even more thorough. Traditional Apps built competitive advantages by monopolizing user behavioral data, whereas the Doubao phone’s system-level permissions enable cross-App data integration. The AI assistant can correlate consumption intent from chat records with pricing data from e-commerce platforms and proactively recommend cost-effective products, effectively collapsing the data-monopoly advantage of any single App.
Behind this conflict lies a struggle for control between an “AI Agent ecosystem hub” and the “App island economy.” The future battlefield has already converged on the user entry point defined by “hardware terminals + on-device AI.” Competition in on-device computing power is well underway. Manufacturers such as Honor and OPPO have explicitly committed to “on-device AI,” with flagship smartphone NPUs expected to exceed 60 TOPS by 2026. Local deployment of larger models will reduce reliance on the cloud while better mitigating data-leakage risks. The contest over interface standards is equally critical. Doubao’s current “simulated user operation” approach easily triggers risk controls; going forward, “Agent-to-Agent” protocols and standardized MCP APIs will become focal points of negotiation. App developers and AI assistant companies must find a balance between efficiency and security. Sensitive scenarios such as finance and healthcare will be central battlegrounds. AI assistants must break through risk-control constraints via tiered authorization mechanisms, while App providers may open selected non-core interfaces in exchange for traffic-sharing arrangements. The seeds of cooperative win-win outcomes are quietly beginning to sprout.
From a long-term perspective, pure technological blockades are unsustainable. The industry will inevitably move from “fragmented confrontation” toward “layered collaboration.” Two major ecosystem camps are gradually taking shape. The “Doubao Alliance,” centered on ByteDance, brings together second-tier smartphone manufacturers and small-to-medium Apps, emphasizing open collaboration. The “traditional giant camp,” composed of Huawei, Xiaomi, OPPO, Vivo, and companies such as Tencent, Alibaba, and Meituan, continues to defend closed-loop ecosystems. At the same time, commercial rules are being redefined. Advertising models will shift from being “exposure-oriented” to “service-oriented.” Brands may pay AI assistants priority recommendation fees, while AI assistants may introduce Pro subscription models priced by task complexity, forming new profit loops. Regulatory intervention will become a key variable, clarifying the boundaries of AI assistant permissions and user data protection requirements—both preventing Apps from unjustifiably blocking AI access and imposing the strictest oversight on sensitive sectors such as finance—to establish clear bottom lines for industry development.
The wall-breaking battle triggered by the Doubao phone is, at its core, a reshaping of industry rules driven by technological innovation. The traffic logic of the mobile internet era is disintegrating, while competition for entry points in the AI era has reached a fever pitch. For enterprises, the choice is stark: either embrace change and seek new growth through open collaboration, or fortify defenses within closed ecosystems to resist disruption. For users, meanwhile, smarter, more efficient, and more convenient human–machine interaction experiences will become an irreversible trend. Amid this disruption and reconstruction, only those who move with the tide of the times will stand firm through the waves of change and witness the arrival of a new internet era.
The NRCap Research Department believes that the sudden emergence of AI Agents represents, in essence, a return of the internet to its original “human-centered” intent. By freeing users from the tedium of price comparisons and the deluge of information, and by returning attention fragmented by content back to life itself, AI enables people to spend more time with family, explore interests, and immerse themselves in real experiences. This is the warmest value of technological progress. However, vigilance is required. New technological revolutions often bring new rule-based challenges. Without effective constraints, AI Agents may create new information barriers. When users become overly dependent on the outputs of a single intelligent agent without understanding its data sources or reasoning logic, they may fall into a new form of “algorithmic cocoon,” even repeating the mistakes of the mobile internet era’s attention wars.
Therefore, the healthy development of AI Agents hinges on the construction of technology and rules grounded in openness and verifiability. As demonstrated by frontier products such as Kaamel Agent, design choices like decision-path visualization, reasoning-chain disclosure, and information traceability allow users to clearly see how AI conclusions are formed and what data supports them, fundamentally eliminating the risks of “black-box decision-making.” This requires industry-wide standards. When AI recommends products, it should clearly label the platforms included in price comparisons and the filtering criteria used. When generating solutions, it must disclose reference sources and weighting logic. When executing actions, interfaces for human intervention and correction must be preserved, ensuring that users always retain final control. Only in this way can technological innovation avoid devolving into a new tool for traffic games, ensure that AI Agents consistently serve user value, and allow the internet to truly become an assistant that empowers life rather than a vortex that consumes energy. Future competition will ultimately be a competition of trustworthy AI, and openness, transparency, and verifiability are the cornerstones of building that trust.