Summary of Key Points
This article uses the 2026 WAIC exhibition as a lens to depict the industrial network that is being shaped by the Agent era. Agents represent the core of current innovation, with both large corporations and startups focusing on their development. However, the entire industry is still in a phase where the direction is clear, but the endpoint remains uncertain, and the paths to achieve it vary. The article explores this through three characteristics (centered around Agents, stimulating the entrepreneurial ecosystem, and open data closed-loop services) and four levels (AgentOS, enhanced services, hardware, and capability ecosystems), illustrating how different players are striving to seize key moments in the Agent era and highlighting the current state of an unsettled industrial order.
Breakdown and Interpretation
1. AgentOS: The New Operating System That Large Corporations Are Fighting Over – Who Will Control the Agent Era?
AgentOS is not a traditional mobile operating system (like Android or iOS); rather, it is a command system tailored for agents. Traditional OSes act as containers for apps, requiring users to find and activate them individually. AgentOS, on the other hand, revolves directly around your intentions. For example, if you say “Book me a high-speed train and hotel for tomorrow in Shanghai,” it can automatically complete the tasks across multiple apps without the need for you to open each one separately.
- Actions of Large Corporations: Honor claims that future phones will be platforms for agents, aiming to reconstruct systems centered on intentions and tasks. Jieye is developing Step AOS, which aims to add a “control layer” to its models to enhance competitiveness by integrating models, agents, and hardware. Nubia’s DouBao phone uses a custom large model at the core, enabling it to handle tasks across apps independently, such as ordering food and booking hotels.
- Startups Are Also Trying: Companies like Xuanjing AI Glasses, Rokid (glasses), and Guangfan Technology (headsets) are attempting to develop AgentOS based on their hardware, though they face greater challenges compared to large corporations. However, there is still room for exploration.
In simple terms, whoever succeeds in creating AgentOS will hold the key to controlling the Agent era, just as Android did in the past.
2. Enhanced Services: Niche Opportunities to Fill the Gaps in AgentOS – Large Corporations Are Not Good at These Detailed Tasks, So Startups Step In
Although AgentOS is powerful, it falls short in handling some detailed needs, providing opportunities for smaller teams:
- Emotional Companionship: Tools like CoolQiqi’s “CookiePi” interactive characters can engage with you, and robots like Babu Bubo can sense your mood and offer suggestions (e.g., suggesting you take a break when you seem tired). These personalized services are difficult for large-scale OSes to provide.
- Memory Optimization: AgentOS also handles memory, but there are shortcomings. For instance, changing agents may result in the loss of previous habits and task contexts. Companies like Zhixia Technology’s MemoPin help ADHD users optimize their memories and identify patterns (e.g., noticing that you’re often pressed for deadlines), while Memmy can consolidate your knowledge and preferences into long-term memory for use by multiple agents.
These services act as “plugins” for AgentOS, filling in the gaps in its functionality.
3. Hardware: Moving From “AI-Driven” to “Agent-Driven” – Devices Are No Longer Just Tools, but Extensions of Agents
Previous hardware (such as phones) were “AI-driven,” enabling AI functions, but in the Agent era, they become “agent-driven,” serving as extensions for agents to perceive the world and interact with you:
- The Phone Is No Longer the Only Hub: Li Yannan from Sequoia suggests that phones are evolving from the center of mobile internet to a core node in the Agent era. Companies like Honor and Jieye are focusing on multi-device collaboration, using the same agent brain to control phones, glasses, and headphones.
- New Hardware Emerging: There are AI-enabled glasses (e.g., Qianwen’s that monitor your vital signs and access third-party services), earbuds (that share data with glasses for enhanced interaction), pendants (e.g., Zhixia Technology’s that record voice commands), and desktop devices (e.g., SlashVibe’s with microphones for issuing instructions and secondary screens to display agent progress). Startups like ClawStage are developing specialized hardware for controlling smart homes, while ViewPoint Transition’s platforms facilitate rapid prototyping for geeks and developers.
These devices are no longer just tools; they become agents’ “eyes, ears, and hands,” helping them perceive your environment and perform tasks.
4. Capability Ecosystem: Tools That Enable Agents to Perform Tasks – Both Niche and Native Tools Are Contributing
For agents to be useful, they need the right tools, which is where the capability ecosystem comes into play:
- Niche Applications Becoming Agent-Friendly: Established players in the mobile internet (e.g., Didi and Alipay) are transforming their services into capabilities that agents can utilize. For example, Didi’s AI Xiaodi can book a red car for you based on your preferences, while Alipay and Lark are making their services more accessible to agents.
- AI-Native Tools: New teams are creating tools specifically for agents. For instance, ChatCut allows you to use voice commands to edit videos and add subtitles, while Open Design provides AI-driven design workflows.
These tools act as “skill packs” that enable agents to perform various tasks, from booking rides to video editing.
5. The Current State of the Industry: The Direction Is Right, but the Winner Is Yet to Be Determined – Everyone Is Competing for Indispensable Positions
The current Agent industry is like a “living map”: the direction (agents becoming core) is clear, but the endpoint and the path to achieve it are uncertain.
- Data Silos: Each player is accumulating user data (e.g., habits and preferences) as a competitive advantage, reluctant to share it, so a unified market is unlikely in the short term.
- Borders Are Blurring: Traditionally, OSes controlled access, apps handled requests, and hardware facilitated interactions. Now, AgentOS aims to directly understand users without going through apps; hardware seeks to become a new entry point, enhanced services want to control key links, and niche applications want to transform into capabilities. Everyone is vying for strategic positions.
The winner will be the one who can solve these issues: open up capabilities while retaining data, access external services, and manage user relationships effectively, enabling seamless collaboration between different devices and agents.
In Conclusion
The Agent era is already on its way, but no one has yet defined the final “rules of the game.” Large corporations and startups are still figuring out how to navigate it. As users, you will soon experience more intelligent and personalized AI assistants.