Abstract:
Programming education is being rapidly transformed by generative AI tools, as these tools powered by large language models (LLMs) have proven to be effective in addressing a wide range of programming-related tasks and have already been made extensive use by students. Leveraging these AI tools in programming is so-called vibe coding. Although vibe coding significantly improves the development efficiency, it has been reported that AI-generated code may contain faults, information leakage, and security vulnerability, and students over-relying on AI could hinder the development of core coding skills and are less likely to succeed in managing large-scale software projects. Moreover, it is reported that top software companies (e.g., Microsoft) begin enforcing more AI tool usage in work, but students lack sufficient guidance in coding with AI.
This initiative aims to support students in learning coding with AI that leverages its rapid prototyping power and collaborates to address the quality issues in a form of “cyber pair programming”. The main goal of this project is to develop techniques and an educational guideline that supports students in excelling at vibe coding. The techniques facilitate students in vibe coding with better interactions with AI by requirement management and software testing. The guideline details warnings, suggestions, and actionable items for students, which can also be adopted into various computational courses in universities to facilitate the teaching and enhance the student’s learning, delivering a new pedagogical approach for AI-assisted programming in the classroom. Echoing the trend of existing workforce, this initiative also equips our students with critical thinkings of AI to prepare them for their future work.
Code:
T0302
Principal Project Supervisors:
Keywords provided by authors:
Start Date:
01 Apr 2026
End Date:
30 Jun 2027
Status:
Ongoing
Result:
- A V-model Pedagogy: This framework will guide educators in teaching coding with AI, specifically within programming education. It provides step-by-step practices that highlight the roles of requirement management and software testing in vibe coding, equipping students to interact effectively with AI tools.
- A Curated Set of Concrete Examples: We will develop and share illustrative cases for lecturers to demonstrate the biases and limitations of AI-generated code, helping students critically analyze and compare AI solutions with human-developed solutions.
- An Integrated VSCode-Based Platform: The project will deliver a user-friendly platform that enables personalized and active learning experiences in vibe coding, incorporating best practices and seamless access to programming exercises.
Impact:
The left panel will contain detailed introduction of the pedagogy and the methodology of vibe coding and provide a set of concrete python program examples. Students can explore different programs along with the guidance of the detailed instructions for students in different proficiency. Following the steps we describe, students can try to interact with the AI agents to complete the designed problems, and then they can execute the program and observe the program outputs responding to our proposed V-model for vibe coding.
Financial Year:
2025-26
Type:
TDG