AI Agents: In-depth Comparison of Baidu's "Xin Xiang" and Byte's "Kuzi Space"

In April, China's AI track welcomed two blockbuster products - Baidu's "Xin Xiang" and Byte's "Kuzi Space". Both are benchmarked against the popular general intelligent agent Manus overseas, but they enter the market with differentiated paths, marking a key turning point for AI from the "dialogue era" to the "execution era". In this competition between big companies, who will win?

1. Product positioning: life assistant VS office partner

1. Baidu "Heart Sound"  

Positioned as a super intelligent entity on mobile phones, it focuses on personal life scenarios, such as travel planning, AI blind dating, and smart chart generation. Through natural language interaction, users can quickly generate GDP charts, children's stories, and even monitor stock fluctuations, achieving "full-process hosting". Its homepage "Inspiration" interface is designed with interesting channels such as "Wow" and "AI Blind Date", intending to attract C-end users through high-frequency life needs.

2. Byte "Button Space"  

It focuses on a general intelligent collaborative office platform, emphasizing "working with agents". It supports the MCP protocol, can call tools such as Feishu documents and Amap, and has built-in vertical field agents such as "User Research Expert" and "Huatai A-share Observation Assistant" to solve workplace needs such as data analysis and market research. The design of its exploration mode and planning mode is more like an "intern", and it needs to interact with users repeatedly to confirm their needs.

2. Functional comparison: scenario-based execution vs. professional collaboration

3. User groups: life-lovers vs. professionals

- Xin Xiang targets individual users and solves daily chores:  

  For example, the "AI Blind Date" channel can match objects by analyzing user portraits; the "Smart Chart" supports the input of natural language to generate data visualization results, which is suitable for groups such as students and housewives.

- Button Space serves workplace users and improves work efficiency:  

  "Huatai A-share Observation Assistant" can generate daily stock market morning reports and combine professional databases to reduce AI illusions; "User Research Expert" can analyze interview records and generate research reports, becoming a "digital colleague" of product managers and operations personnel.

IV. Technical Path: Large Model Integration VS Ecosystem Openness

- Xin Xiang relies on Baidu's big model and multi-agent collaborative technology, emphasizing "closed-loop solutions", but is slightly single in tool chain integration.

- Kouzi Space breaks the ecological barriers through the MCP protocol: users do not need to configure APIs or codes, and can call tools such as Feishu and Gaode with one click, and even generate HTML web pages with interactive effects. Its Doubao model supports "thinking and searching at the same time", combined with an ultra-low latency of 20 milliseconds, which is more suitable for complex task processing.

5. Commercialization Challenges: Breaking through Homogeneous Competition

Both face two major challenges:  

1. Technology homogeneity: core capabilities rely on large models and tool chain integration, and no absolute barriers have been formed;  

2. Cultivate payment habits: Users have limited acceptance of “paid hosting services” and need to explore models such as subscriptions and enterprise-level services.  

Currently, Kouzi Space is testing the market through "limited-time free" expert agents, while Xinxiang may rely more on advertising or member value-added services.

6. Industry enlightenment: The final battle of AI lies in "scenario definition"

Whether it is Xin Xiang's "full life hosting" or Kou Zi Space's "office partner", the essence is to tap into users' unmet task needs. For example:

- Xinxiang's "AI blind date" directly hits the social pain points, but the issues of privacy and matching accuracy need to be resolved;  

- The "Stock Analysis Assistant" of Kouzi Space relies on professional data sources and needs to reduce execution failures caused by API permissions.  

In the future, whoever can seamlessly integrate AI capabilities into high-frequency scenarios will be able to seize the initiative in the blue ocean of intelligent entities.

The competition between Baidu and ByteDance is not only a competition of technical strength, but also a deep insight into user needs. Xinxiang lowers the threshold for AI use with a "lightweight life assistant", and Kouzi Space reshapes office processes with a "professional tool ecosystem". 2025 is the first year of commercial use of intelligent bodies, and the ripples of this revolution have just spread - perhaps soon, each of us will have an "AI partner" online 24 hours a day.