Moonshot AI's Yang Zhilin: One-Million-Token Model Signals New AI Paradigm

The pace of AI evolution may soon be directed by AI itself.
On March 25, at the 2026 Zhongguancun Forum Annual Meeting, Yang Zhilin, founder of Yue Zhi An, delivered a keynote speech. He predicted that within the next year or so, the research and development methodology of artificial intelligence will undergo a fundamental shift, transitioning the role of human researchers from "hands-on builders" to "resource orchestrators."
A New R&D Paradigm: From "Human-Brain-Led" to "Token-Driven"
Yang Zhilin noted that AI development is entering a new, AI-led phase. Within this paradigm, the working style of researchers will change profoundly:
Tokens as Core Resources: In the future, each researcher will be equipped with massive amounts of AI Tokens. These tokens will transcend their role as mere conversation units, becoming fundamental productivity resources.
AI-Powered Self-Exploration: With abundant tokens, AI can assist researchers in synthesizing novel tasks, creating new training environments, and even defining optimal "reward functions," enabling autonomous exploration of innovative neural architectures.
The Efficiency Singularity: "Self-Driving" AI Research and Development
This transformation means AI iteration will no longer be entirely constrained by the cognitive limits and output of human experts.
Exponential Acceleration: As AI begins to steer the research process itself, the pace of technological advancement is poised for exponential growth.
Building an Ecosystem: Yue Zhi An expressed its commitment to collaborating closely with the open-source community to collectively push the frontiers of intelligent technology and foster a more dynamic ecosystem.
Industry Landscape: Large Models Wade into the "Agentic Execution" Era
Concurrent with Yang Zhilin's address, the industry is witnessing a significant surge in trends around "Agent (Intelligent Entity)" capabilities.
Tencent: Its AI assistant, Crab (WorkBuddy), received updates that significantly enhance its utility in office productivity scenarios.
Step Star: The cloud-based AI assistant StepClaw, built upon the OpenClaw framework, has also officially launched.
Conclusion: When AI Researches AI
The shift from "teaching AI to communicate" to "empowering AI to conduct R&D" represents more than a tool upgrade; it is a transformation of production relations. In Yang Zhilin's vision, within the leading laboratories of the future, a key metric for a researcher's prowess may become their efficiency in directing AI Tokens to drive innovation.
As AI R&D enters this "self-driving" stage, the path to Artificial General Intelligence (AGI) may be shorter than we anticipate.
Related article
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage
California AV Compliance: A New Era of Tickets, Geofences, and 1M Miles
Guident operates an AuveTech shuttle in South Florida, managing a four-mile route in West Palm Beach and a one-mile route in Boca Raton using its remote monitoring technology. | Credit: GuidentCalifornia is redefining the regulatory landscape for dri
Related Special Topic Recommendations
Comments (0)
0/500

The pace of AI evolution may soon be directed by AI itself.
On March 25, at the 2026 Zhongguancun Forum Annual Meeting, Yang Zhilin, founder of Yue Zhi An, delivered a keynote speech. He predicted that within the next year or so, the research and development methodology of artificial intelligence will undergo a fundamental shift, transitioning the role of human researchers from "hands-on builders" to "resource orchestrators."
A New R&D Paradigm: From "Human-Brain-Led" to "Token-Driven"
Yang Zhilin noted that AI development is entering a new, AI-led phase. Within this paradigm, the working style of researchers will change profoundly:
Tokens as Core Resources: In the future, each researcher will be equipped with massive amounts of AI Tokens. These tokens will transcend their role as mere conversation units, becoming fundamental productivity resources.
AI-Powered Self-Exploration: With abundant tokens, AI can assist researchers in synthesizing novel tasks, creating new training environments, and even defining optimal "reward functions," enabling autonomous exploration of innovative neural architectures.
The Efficiency Singularity: "Self-Driving" AI Research and Development
This transformation means AI iteration will no longer be entirely constrained by the cognitive limits and output of human experts.
Exponential Acceleration: As AI begins to steer the research process itself, the pace of technological advancement is poised for exponential growth.
Building an Ecosystem: Yue Zhi An expressed its commitment to collaborating closely with the open-source community to collectively push the frontiers of intelligent technology and foster a more dynamic ecosystem.
Industry Landscape: Large Models Wade into the "Agentic Execution" Era
Concurrent with Yang Zhilin's address, the industry is witnessing a significant surge in trends around "Agent (Intelligent Entity)" capabilities.
Tencent: Its AI assistant, Crab (WorkBuddy), received updates that significantly enhance its utility in office productivity scenarios.
Step Star: The cloud-based AI assistant StepClaw, built upon the OpenClaw framework, has also officially launched.
Conclusion: When AI Researches AI
The shift from "teaching AI to communicate" to "empowering AI to conduct R&D" represents more than a tool upgrade; it is a transformation of production relations. In Yang Zhilin's vision, within the leading laboratories of the future, a key metric for a researcher's prowess may become their efficiency in directing AI Tokens to drive innovation.
As AI R&D enters this "self-driving" stage, the path to Artificial General Intelligence (AGI) may be shorter than we anticipate.
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage





Home






