For the past decade, the success of consumer technology platforms has been measured by a single, ubiquitous metric: Daily Active Users (DAU). Now, Baidu is attempting to rewrite the scorecard for the artificial intelligence era.
At its annual Create 2026 developer conference in Shenzhen, the Chinese search giant declared that the industry has officially entered the “Agent Era.” To mark the transition, Baidu co-founder and CEO Robin Li proposed a new key performance indicator that he argues will define the next generation of technology companies: Daily Active Agents (DAA).
The shift in terminology is more than just marketing spin; it reflects a fundamental change in how Baidu envisions the future of human-computer interaction. According to the company’s official press release, the focus is moving away from passive users consuming content toward active, autonomous software entities performing tasks on behalf of humans. Li predicted that global DAA could eventually surpass 10 billion, a figure he positioned as the agent-era equivalent of the mobile internet’s billions of daily active users.
A Three-Layer Transformation
During his keynote, Li outlined what he described as a three-layered transformation underway in the AI industry. At the first layer, agents are evolving from passive responders into autonomous executors that continuously learn from their environment. At the second layer, individual users are evolving from ordinary consumers into “super individuals” who learn to coexist with and direct AI systems. At the third layer, enterprises are evolving from human-to-human collaboration structures into what Li called “mixed human-agent formations,” or “super-organizations.”
This framework underpins Baidu’s product strategy. Rather than building a single flagship AI assistant, the company has developed a portfolio of specialized agents designed to address distinct use cases across consumer and enterprise markets.
The New Agent Portfolio
The centerpiece of the Create 2026 announcements was the launch of DuMate, Baidu’s general-purpose agent. The mobile edition of DuMate was released at the conference, featuring full real-time synchronization between mobile and PC, 24/7 availability, and the ability to read screens, operate software, process files, and connect to business systems. Li described DuMate as capable of handling the full range of tasks a user might delegate to a capable personal assistant, and the company claimed it achieves state-of-the-art performance on multiple international agent benchmarks.
Also launched was Miaoda, Baidu’s coding agent, which targets both individual developers and enterprise teams. Miaoda is designed to enable “vibe coding,” the generation of functional applications through natural language prompts, without requiring the user to write code directly. In a striking demonstration of the tool’s capabilities, Li noted that 90 percent of Miaoda’s own code was generated by Miaoda itself. “Coding agents like Miaoda enable people with no coding knowledge to generate applications directly; the barrier is disappearing,” Li said. “At the same time, development costs are being compressed to near zero. The software industry is being redefined.”
A third new product, Baidu Yijing, is a full-scenario digital human platform rebranded from the company’s earlier Huiboxing service. It supports live streaming, video production, and real-time interactive experiences across 12 languages with native-level lip-sync accuracy. “A digital human is simply an agent you can see,” Li explained. “Equipped with voice, facial expressions, and gestures, a digital human is more expressive and more trustworthy.”
Finally, Baidu unveiled Famou Agent 2.0, an upgraded version of its self-evolving industrial agent. The new version focuses on three high-value enterprise use cases: production scheduling, process optimization, and logistics planning. Baidu claimed a 10.21 percent performance improvement over an already highly optimized baseline, and noted that Famou Agent 2.0 topped MLE-Bench, winning 9 of the 15 hardest tasks and outperforming agents built on other leading models.
The Competitive Stakes
Baidu’s aggressive push into the agent space comes at a critical time for the company. While it was an early pioneer in Chinese AI research, it has faced intense competition from agile startups like Moonshot AI and DeepSeek, as well as the massive distribution advantages of Tencent’s WeChat ecosystem.
As we reported earlier today, Tencent’s WorkBuddy has become the most widely used AI agent in China, thanks to its integration with WeChat Work, which has enabled massive scale. Baidu lacks a comparable consumer messaging platform, forcing it to rely on its search dominance and enterprise cloud relationships to drive agent adoption.
The success of the “Agent Era” strategy will depend heavily on Baidu’s ability to attract a vibrant developer community to its new agent platforms. If Baidu can successfully establish DAA as the new industry standard, it could reshape how investors evaluate AI companies, shifting the focus from raw model capabilities to practical, task-oriented utility. For now, the metric serves as a bold declaration of intent from a company determined to define the terms of the next phase of the AI revolution.