AWS Logo
Menu
First time with Amazon Q: Game app for hackathon

First time with Amazon Q: Game app for hackathon

I love participating in hackathons, working with AI has drastically improved my efficiency at getting the project complete on time. This time, we tried to use the AI to build apps on another platform (reddit). Work well, but requires a lot of hand-holding.

game-challenge
choong
Published Dec 30, 2024

Introduction

Hi everyone, thanks to the hackathon challenge from Devpost, I wasn't aware of the code assist that Amazon has offered with the Amazon Q developer. With this, I get to work on some little projects while experiencing what Amazon's code assist offers.
Before that, let me share a little about my game project called "Guess the Pokemon". To be fair, was already building a game project for another hackathon on Devpost. I also made a rule for myself to use Amazon Q only.
TLDR. The result has been pretty impressive, wasn't what I was expecting. Let me break down those experiences later.
  • Here is the link to the demo: https://www.youtube.com/watch?v=GVQfD_3Zozg
  • Here is the link to the app: https://developers.reddit.com/apps/guess-pokemon

Experience

Aside from not being able to automatically update the change to the code base, I think the rest are consider pretty impressive for a free tier usage.
Here are some of the few example problems I just let the AI to solve.
  1. Generating the Pokemon silhouette based on the provided image.
    1. This highly depends on the quality of the image I provided. Initially, I just did a random search (image with background). It can't generate a logic that has good edge detection.
    2. Therefore, I resorted to getting the Pokemon with a transparent background. (this time the detection is much better).
  2. Adjusting the CSS styling, size, and theme (dark/light).
  3. Input guess with adjustable typo tolerance.
Adjusting the style and overlay to provide the iconic layout
Given the knowledge of the internet, each Pokemon has its number and generation name. The autocomplete was pretty impressive to detect and changed my code implementation based on their recommendation.

After everything was done, I tried to get the AI to refactor the CSS since it looked a little messy. Then again, I think it lacks the context of the whole project to do proper refactoring. Which I think, could be improved by referring to other VS code extension implementations.

Conclusion

Overall, the whole process still requires a lot of hand-holding to guide the AI to fix problem, and provide technical information for it to proceed. Eg. making use of the Devvit API, external data source(pokedex), etc.
Nevertheless, a drastic improvement compare to 1 year ago for software development.
 

Comments