On August 21, with the support of the People's Education Audio-Visual and Digital Press, Xiaoyuan Learning Machine hosted an official launch event for the "Primary and Secondary School Textbook Learning Agent" in Beijing, where both parties signed a cooperation agreement. The agent was spearheaded by the People's Education Audio-Visual and Digital Press, leveraging high-quality textbook content resources and Xiaoyuan's self-developed Yuanli LLM as its joint cornerstone. Xiaoyuan AI Learning Machine serves as the first distribution channel, bringing authoritative textbook content directly into students' self-directed learning scenarios in the form of an AI-powered learning agent.
Policy and industry momentum are now moving in lockstep. This April, the Ministry of Education and four other government departments jointly issued the "AI + Education" Action Plan, which explicitly calls for the development of intelligent learning companions and the integration of educational large models and agent tools. Riding this policy tailwind, agent products built around authoritative textbook content that systematically serve learning scenarios have become a critical missing piece for the education industry—and leading learning hardware makers are racing to seize this entry point.
From a product design perspective, this system goes far beyond "tool-level" AI. The "AI companion learning" mode transforms static textbook text and imagery into visualized, AI-powered dynamic textbooks, offering features such as tap-to-read read-alongs, recitation assessment, and AI Q&A to tackle the "willingness to learn" challenge. The "AI guided learning" mode, meanwhile, deploys an AI teacher that maps learning paths by academic stage and subject according to curriculum standards and textbook compilation logic, using one-on-one interactive explanations to address the "how to learn" challenge. Together, the two complementary modes support students across a wide spectrum of self-learning abilities.
Pilot data further validates the design. Since the pilot program began in March 2026, the agent has attracted more than 100,000 primary and secondary school students. Among sustained users, Chinese language knowledge point mastery improved by 33.6% and English by 24%, with speaking participation rates reaching 81.4% for Chinese and 61.7% for English. Parents' daily accompanying-study time dropped by 20 minutes. Still, whether educational AI can operate reliably and sustainably in real classroom scenarios—and how content safety boundaries are safeguarded—remains the defining test for large-scale adoption.
At its core, an education agent is about bringing authoritative content, model capabilities, and learning scenarios into a single, operable system. StarWar Cloud focuses on GPU computing power platforms and the AIedulab training platform, helping enterprises unify the management of teaching content, model invocation, and learning scenarios. By providing a stable computational foundation and engineering-grade support, StarWar Cloud moves AI education training from occasional demonstrations to everyday, normalized practice.
Looking at broader industry trends, textbooks are evolving from "static texts" into multimodal, highly interactive dynamic learning systems. Competition in educational AI is likewise shifting from raw model-parameter comparisons to content quality and contextual understanding of the learning process. As authorized publishing institutions and education technology companies deepen their collaboration, the closed loop between "what to learn" and "how to master it" will come to define the industry's competitive moats.
The larger significance of this collaboration is that it demonstrates how educational AI is not simply about content processing or feature stacking—it is about enabling high-quality content to genuinely enter the learning process while respecting both publishing norms and pedagogical principles. Textbook learning is moving from "finishing the lesson" to "truly understanding the material," and the next chapter of education digitalization is being redrawn by agents of this kind.