The widespread adoption of large language models (LLMs) has presented notable challenges to traditional paradigms of experimental computer teaching in universities. Traditional educational objectives and teaching models no longer align with the talent development needs of the LLM era. Recent studies have focused on the application of LLMs in experimental computer teaching. However, existing research often lacks an in-depth analysis of the limitations of LLMs and fails to sufficiently explore the evolving competency requirements for computer professionals or human–artificial intelligence (AI) collaboration patterns. Therefore, this paper thoroughly investigates the new competency requirements of the current job market for computer science students, redefines the goals of experimental computer teaching, and constructs a novel experimental computer teaching system centered on human–AI collaboration. The goal is to develop individuals equipped with innovative thinking, complex problem-solving skills, and the ability to collaborate with AI to meet the opportunities and challenges brought about by the LLM era.

To gain insight into industry-aligned competencies, a questionnaire-based survey was conducted targeting recent graduates from Beihang University who were employed in the computer industry. The survey explored the practical use of LLMs in their work, including how LLMs were applied in programming tasks, their strengths and limitations, and the extent to which programmers could be substituted. Based on the survey results, this paper analyzes the emerging competency requirements in the job market and clarifies the evolving goals of experimental computer teaching. Furthermore, it examines the opportunities and challenges caused by industrial transformation. Based on the analysis, this paper systematically designs corresponding teaching content and instructional models. Specifically, it proposes a project-based experimental teaching system supported by LLMs and designs a comprehensive project for intelligent robots driven by LLMs, thereby providing a detailed explanation of how to implement human–AI collaborative teaching.

The novel experimental computer teaching system was officially implemented at Beihang University and was supported by an integrated teaching and research platform named “Xiji”. This system establishes a tripartite collaborative teaching model involving teachers, LLMs, and students. Based on this platform, a comprehensive LLM-driven project on intelligent robots has been adopted as a core experimental component of AI training for third- and fourth-year undergraduates from the School of Computer Science and Engineering, the School of Artificial Intelligence, and the School of Mechanical Engineering and Automation. It is also integrated into graduate-level courses such as Intelligent Control of Robots. Practical results indicate that the project has achieved positive outcomes in real-world teaching, facilitating the deep integration of LLMs effectively with engineering practice and developing students’ competency in human–AI collaboration and comprehensive innovation.

This paper presents a systematic review and in-depth analysis of the current applications of LLMs in computer science, the evolving demands of the job market under new competency structures, and the advantages and innovative practices of project-based learning. In the future, universities should actively promote a structural transformation of experimental computer teaching, shifting from the traditional teacher–student binary relationship to a tripartite collaborative model involving teachers, LLMs, and students. By introducing LLM-assisted project-based learning, students can be guided effectively to develop capabilities in active exploration, deep thinking, and systematic innovation. This teaching model provides a practical and feasible pathway to support the transformation and advancement of talent cultivation in universities in the LLM era.