Xiaomi Kuaiyingyong's head Dong Hongguang leaves to start a business, aiming to create an intelligent agent for AI hardware. | Exclusive from Zhiyong.

Written by | Wang Fangyu

Edited by | Su Jianxun

"Intelligent Emergence" has exclusively learned that Dong Hongguang, the person in charge of Xiaomi's Quick App and a former employee of the top 100, has recently left to start his own business. The direction of his entrepreneurship is AI applications based on hardware products.

Currently, there is a wave of AI hardware entrepreneurship at home and abroad, including AI toys, AI companion robots, AI glasses, AI headphones, etc. According to informed investors, what Dong Hongguang wants to do is the software (also called Agent or Intelligent Agent) installed on the above AI hardware.

Public information shows that Dong Hongguang joined Xiaomi Group in 2010 and is one of the top 100 employees of Xiaomi. He has successively served as the person in charge of the MIUI system framework, the person in charge of the Quick App ecosystem business, and the person in charge of the self-developed operating system.

During his time at Xiaomi, Dong Hongguang was initially responsible for the construction of the Xiaomi Theme Store, participated in the underlying technology research and development of MIUI, and then incubated the Quick App Ecosystem Alliance and the self-developed operating system business. He has been responsible for the operating system and application ecosystem and has relatively rich software experience.

Several investors close to Dong Hongguang told "Intelligent Emergence" that for potential investment institutions, Dong Hongguang's main advantage is the "High-P Team". Including Dong Hongguang, almost all team members are high-level technical professionals from large companies, with work experience in large companies such as ByteDance, Alibaba Cloud, WeChat, Sogou, and Motorola.

Under the wave of generative AI, multimodal human-computer interaction technology is gradually maturing, which may bring potential transformative opportunities to the intelligent hardware industry. Therefore, the AI hardware track is attracting more and more attention and investment. At a time when venture capital is cooling down, AI hardware is also one of the few areas that can frequently obtain financing.

Because these AI hardware are still in the early stage of development, and the software ecosystem is not yet perfect, there is a demand gap for software and the ecosystem.

A typical scenario is that users wear headphones, glasses, or talk to toys, and the AI Agent can provide different functions such as information query, object recognition, real-time translation, travel planning, restaurant recommendation, English learning, and emotional communication.

In the long term, with the improvement of computing power and the optimization of AI models, AI hardware may become the next super hardware and independent entrance after mobile phones, and the native applications of AI hardware have the opportunity to grow and become a new generation of "killer" applications.

The interaction methods of AI hardware represented by AI headphones and AI glasses are fundamentally different from those of mobile phones. The core interaction method of AI glasses and headphones is mainly voice. Users complete operations (such as "take a photo" and "translate") through natural language instructions without the need for a touch screen or complex gestures.

Traditional applications are where people find services and interact passively through the GUI of software and apps; while the AI Agent on AI hardware is a voice interaction between people and the intelligent agent, which is more like an active personal assistant, which puts forward new requirements for the development of AI Agents. This transformation is a new market opportunity for entrepreneurs represented by Dong Hongguang.

The founder of an AI glasses startup told "Intelligent Emergence" that the application development cost of AI glasses is very low. If a suitable scenario can be found, it is possible for one or two developers to create the agent.

However, the types of AI software are complex and diverse, such as tool types, entertainment types, education types, etc. The audience groups, market space, and user stickiness of different types of software are very different. Like WeChat and Douyin in the era of the mobile Internet, to create a "killer" native application for AI hardware undoubtedly requires a lot of exploration and trial and error.

An investor said that domestic C-end users lack the habit of paying subscription fees for software, and it is easier to run the business model through the B-end channel in the initial stage; while in the overseas market, it is possible to cooperate with AI hardware companies to provide applications and collect C-end subscription fees, which is more friendly to startups.