Building "Truth or Dare" with Amazon Q
My experience in AWS Hackathon
Published Jan 4, 2025
I started developing the classic party game "Truth or Dare" as part of AWS Hackathon. Throughout this project, I utilized various tools, Amazon Q, Amplify and Bedrock models to assist in the development process. Here's a detailed account of my experience:
About Myself:
- 15 years of web development experience
- Mid-level skills in React and AWS
- Using Github Copilot since launch
- 15 years of web development experience
- Mid-level skills in React and AWS
- Using Github Copilot since launch
Code Generation and Assistance
- Amazon Q: The majority of the code—approximately ~90%—was generated using Amazon Q, Most of the files are fully generated multiple times with different prompts.
- Debugging and Human Input: The remaining ~5% of the code involved online debugging and incorporating human input. While Amazon Q was highly effective, certain complex scenarios required manual intervention and expertise to resolve issues.
- Additional Tools: I also utilized ChatGPT and GitHub Copilot for about ~5% of the development. These tools offered alternative perspectives and solutions, complementing the assistance provided by Amazon Q.
Challenges Encountered
While Amazon Q proved to be a valuable asset, several challenges emerged during the development process:
- Extension Switching: When switching to another extension, Amazon Q would move to the top of the chat, disrupting the workflow. This behavior necessitated manual adjustments to return to the previous context.
- Need More Actions: The absence of a "Create file" action within Amazon Q required manual file creation, adding an extra step to the development process.
- Policy Errors: Certain valid responses were erroneously flagged due to Amazon's responsible AI policy, leading to unnecessary debugging and clarification.
- Cursor Insertion: The "Insert at cursor" feature did not function as expected within the terminal, hindering seamless code insertion.
- Inline Completion: Inline code completion was less effective compared to GitHub Copilot, occasionally requiring manual code adjustments.
Positive Aspects
Despite the challenges, several positive aspects of using Amazon Q stood out:
- Detailed Responses: Amazon Q provided longer, more comprehensive answers compared to GitHub Copilot, often including full file implementations. This depth of information was particularly beneficial for complex coding scenarios.
- AWS-Specific Suggestions: The tool offered up-to-date code suggestions, especially related to AWS services, Including configuration and best practices.
In conclusion, Integrating AI-powered tools like Amazon Q can substantially improve efficiency and code quality in web development projects.