Li Yanhong: "Self-Evolution" Spans Agents, Individuals, and Enterprise Organizations

Background and Context

On May 13, Li Yanhong, the founder of Baidu, delivered the keynote address at the opening ceremony of the Create 2026 AI Developer Conference. In this speech, he systematically articulated a new paradigm for the evolution of artificial intelligence, centering on the concept of "self-evolution." This framework is not merely a technical roadmap but a comprehensive sociological and economic theory that redefines the relationship between technology, individuals, and organizations. Li Yanhong posited that the AI industry has moved beyond the initial phase of large model parameter wars and entered a critical stage of application落地 (landing) and ecological reconstruction. The core of his argument is that AI is no longer just a tool to be called via API or embedded as a feature, but has become an autonomous element capable of co-evolving with humans.

The keynote outlined three distinct layers of this self-evolution, which together form a triad of transformation. First is the self-evolution of AI agents, which are shifting from passive response mechanisms to active entities that learn from their environment and execute tasks autonomously. Second is the evolution of human individuals, where ordinary people leverage AI to become "super-individuals," acquiring new skills to collaborate and coexist with intelligent systems. Third is the self-evolution of enterprise organizations, which are transitioning from traditional human-to-human division of labor to "super-organizations" composed of mixed formations of humans and AI agents. This triad represents a shift from tool-assisted productivity to deep integration, marking a fundamental change in how value is created in the digital economy.

This announcement serves as a strategic signal from Baidu, positioning the company not just as a technology provider but as a definer of new production relations in the AI era. By framing the future around these three dimensions, Li Yanhong is attempting to steer the industry conversation away from raw computational power and towards practical utility, ecosystem depth, and organizational agility. The Create 2026 conference, therefore, acts as a barometer for the broader AI industry, indicating a consensus among tech leaders that the next phase of growth depends on the seamless integration of autonomous agents into daily workflows and business structures.

Deep Analysis

The first dimension of self-evolution, the transformation of AI agents, represents a significant leap from concept validation to scalable deployment. Traditional AI applications have largely relied on prompt engineering, operating in a reactive mode where users input commands and models generate outputs. In contrast, the "self-evolving" agents described by Li Yanhong possess the capability to perceive their environment, plan actions, and execute tasks independently. This shift is underpinned by advancements in multimodal large model reasoning and the refinement of long-term memory modules. These technical improvements allow agents to learn patterns from unstructured data, decompose complex tasks, and utilize toolchains to close the loop on execution. Essentially, agents are evolving to accumulate experience in dynamic environments, much like human workers, thereby reducing the need for constant human oversight.

The second dimension focuses on the emergence of the "super-individual." This concept does not refer to biological enhancement but rather to a massive expansion of cognitive leverage. By mastering new paradigms of collaboration with AI, ordinary individuals can overcome skill gaps and achieve output levels that previously required entire teams. This transformation requires individuals to shift their role from executors to commanders. The critical skills for this new era include advanced prompt engineering, logical task decomposition, and the ability to evaluate AI-generated results. The super-individual is empowered to handle complex, multi-step projects alone, fundamentally altering the productivity potential of the individual worker and challenging traditional notions of labor specialization.

The third dimension involves the restructuring of enterprise organizations. Li Yanhong describes a transition from hierarchical, human-centric divisions of labor to "mixed formations" where humans and AI agents work together. In this new organizational model, AI agents are integrated as formal components of the workforce, taking on repetitive, data-intensive, and routine tasks. Human employees, freed from these burdens, can focus on creative endeavors, strategic decision-making, and emotional connection. This hybrid structure not only enhances operational efficiency but also reshapes the cost structure and decision-making chains of companies. Organizations become more agile and flat, capable of scaling operations without the linear increase in headcount associated with traditional growth models. This evolution suggests that the future of work is not about humans replacing agents or vice versa, but about creating synergistic teams that maximize the strengths of both.

Industry Impact

For Baidu, the articulation of the "self-evolution" framework is a strategic move to secure a high ground in the competitive AI landscape. By shifting the focus from model parameters to the richness of the agent ecosystem and the depth of implementation scenarios, Baidu aims to differentiate itself from competitors. The company is leveraging its Wenxin large models in conjunction with an agent platform to build infrastructure that allows developers to create, deploy, and manage agents at a low cost. This strategy is designed to consolidate Baidu's dominant position in the Chinese AI ecosystem, creating a network effect where more developers and users drive further adoption and refinement of the platform. The emphasis on practical application over theoretical capability positions Baidu as a pragmatic leader in the industry.

This trend has profound implications for other tech giants such as Alibaba, Tencent, Microsoft, and Google. The competition is no longer solely about who has the most powerful models or the most computing power, but about who can best enable users to become super-individuals and who can provide mature tools for managing mixed human-agent teams. Companies that fail to adapt to this paradigm risk losing relevance as their platforms become less attractive to developers and enterprises seeking tangible productivity gains. The pressure is on to develop standardized protocols for agent interaction, reduce collaboration costs between different agents, and provide robust management tools for hybrid workforces. This shift is forcing the entire industry to rethink its product roadmaps and go-to-market strategies.

For the broader workforce and developer community, the rise of super-individuals and mixed organizations presents both significant opportunities and substantial risks. Professionals who can quickly adapt to human-AI collaboration and learn to harness the power of agents will gain a competitive advantage in the labor market. They will be able to accomplish more with less, commanding higher value for their strategic and creative contributions. Conversely, those who cling to traditional workflows and resist integrating AI into their daily tasks may face marginalization. This dynamic is already beginning to reshape hiring practices and skill requirements across industries, with a growing demand for roles that emphasize AI literacy, critical thinking, and complex problem-solving over rote execution.

Outlook

Looking ahead, the maturation of agent technology is expected to lead to the proliferation of specialized agents in vertical domains. These agents will play crucial roles in fields such as medical diagnosis, legal assistance, and software development, where precision and domain-specific knowledge are paramount. Key developments to watch include the emergence of standardized interaction protocols for agents, which will facilitate seamless collaboration across different platforms and reduce integration friction. Additionally, it will be interesting to observe whether enterprises begin to pilot "mixed formation" organizational structures on a large scale, potentially giving rise to new management disciplines focused on human-agent team dynamics. The educational sector will also need to adapt, updating curricula to cultivate the next generation of talent with strong AI collaboration skills.

However, the path to full realization of this self-evolution framework is fraught with challenges. Questions regarding the boundaries of agent autonomy, the attribution of responsibility in human-AI interactions, and the ethical implications of automated decision-making remain unresolved. These issues require a balanced approach involving technical iteration, social norms, and regulatory frameworks. Companies must establish governance mechanisms for human-AI collaboration to ensure safety, fairness, and accountability. As AI becomes more integrated into critical business processes, the need for robust oversight and ethical guidelines will only intensify.

Li Yanhong's theory of triple evolution provides a clear roadmap for the industry, but its successful implementation depends on the collective effort of technology providers, enterprises, and individuals. The transformation of productivity in the AI era is irreversible, and those who actively embrace this self-evolution will be best positioned to thrive in the new competitive landscape. Baidu's signal from the Create 2026 conference indicates a strategic intent to lead this transformation, defining the norms and standards for the next generation of digital work. This move is likely to have a lasting impact on the innovation direction of the Chinese internet industry and potentially influence global AI development trends. The focus is shifting from building smarter models to building smarter ecosystems, where technology and humanity evolve together in a symbiotic relationship.