Large language models have catalyzed significant changes in software development through AI-powered coding assistants. This study examines prompt design strategies and features of three widely-used tools: GitHub Copilot, ChatGPT, and Amazon CodeWhisperer. Through systematic benchmarking using algorithmic problems, API integration tasks, and debugging scenarios, we quantify performance differences across accuracy (71-92%), executability (71-94%), and security vulnerability rates (3-12%). GitHub Copilot demonstrates strongest algorithmic performance (92%), while CodeWhisperer excels in API integration (88%) with lowest security vulnerability rate (3%). We provide evidence-based tool selection guidelines and a practical prompt engineering checklist for developers and educators.
목차
Abstract 1. Introduction 1.1 Overview of AI-assisted programming and its growing role 1.2 Importance of prompt engineering 1.3 Research objectives and methodology 2. Functional Overview of AI Coding Tools 2.1 GitHub Copilot 2.2 OpenAI ChatGPT 2.3 Amazon CodeWhisperer 3. Comparative Analysis of Prompt Efficiency 3.1 Benchmark Methodology and Experimental Design 3.2 Quantitative Performance Results 3.3 Detailed Scenario Analysis 3.4 Prompt Structure Impact Analysis 3.5 Language and Framework Adaptability 4. Discussion and Practical Implications 4.1 Language and Framework Adaptability 4.2 System-Level Context and Integration Patterns 4.3 Effective Prompt Patterns 4.4 Implications for Development Practice 4.5 Educational Considerations 5. Conclusion and Recommendations 5.1 Summary 5.2 Evidence-Based Tool Selection Guidelines 5.3 Prompt Engineering Checklist 5.4 Study Significance Acknowledgments References
키워드
AI coding assistantsprompt engineeringcode generationprompt designcontextual code completion
저자
Yo-Seob Lee [ Professor, Dept. of Smart Contents, Pyeongtaek University ]
Corresponding Author