What's Happening?
Junyang Lin, the former technical lead of Alibaba's Qwen project, has shifted his focus from hybrid thinking models to agentic thinking in AI. Lin, who stepped down from his role on March 3, 2026, now operates as an independent researcher. In a recent
talk, Lin discussed the evolution of the Qwen model family, highlighting the transition from training models to training agents. He emphasized the limitations of hybrid thinking, which combines step-by-step reasoning with near-instant responses, and advocated for agentic thinking, which involves AI systems that can plan, act, and adapt based on environmental feedback. Lin's presentation detailed the architecture of the Qwen3 model and its capabilities, including multilingual support and dynamic thinking budgets. He argued that agentic thinking is essential for AI systems to effectively interact with their environments and perform complex tasks.
Why It's Important?
The shift from hybrid to agentic thinking in AI development represents a significant change in how AI systems are designed and evaluated. Agentic thinking focuses on the ability of AI to act and adapt in real-world environments, which is crucial for applications such as autonomous vehicles, robotics, and complex decision-making systems. This approach could lead to more robust and versatile AI systems that can handle a wider range of tasks and adapt to new challenges. The emphasis on agentic thinking also highlights the need for advanced infrastructure to support AI training and deployment, including decoupled training and inference processes. This shift could impact various industries by enabling more efficient and effective AI solutions, potentially leading to advancements in automation, productivity, and innovation.
What's Next?
As the AI field moves towards agentic thinking, researchers and developers will need to focus on creating environments that allow AI systems to learn and adapt effectively. This includes developing infrastructure that supports the decoupling of training and inference, as well as ensuring high-quality environments for AI to interact with. The transition to agentic thinking may also require new evaluation metrics that prioritize task success and adaptability over traditional benchmarks. Companies and research institutions may need to invest in new technologies and methodologies to support this shift, potentially leading to collaborations and innovations in AI development. The broader adoption of agentic thinking could drive significant changes in how AI is integrated into various sectors, influencing future research and development priorities.
Beyond the Headlines
The move towards agentic thinking in AI raises important ethical and practical considerations. As AI systems become more autonomous and capable of interacting with their environments, questions about accountability, transparency, and safety become increasingly relevant. Ensuring that AI systems make decisions that align with human values and societal norms will be a critical challenge. Additionally, the development of agentic AI may require new regulatory frameworks to address potential risks and ensure responsible deployment. The focus on agentic thinking also underscores the importance of interdisciplinary collaboration, as insights from fields such as cognitive science, robotics, and ethics will be essential in shaping the future of AI.
