Anthropic’s Claude 3.5 Sonnet came out a few weeks back and I mentioned its new features to my wife over breakfast. My daughter asked “can it create a game?” I said, “probably- do you want to make a game with me?”
This project is the result of pair programming with Claude and my 3rd grader turned Product Owner. What a great way to dive into Claude’s latest features!
The Wow:
The big takeaway is that Clause 3.5 Sonnet is an imperfect, yet amazing tool. I’ve used gamification with previous teams to teach them how to use AI safely and securely. This is a micro-example of that that’s not under NDA.
Paired programming in this space with the big three LLM’s can help skyrocket development, empower and educate less skilled teams and interns, and in the right hands, help build faster MVP’s and solutions getting the mundane out of the way.
What We Loved:
1. Instant MVP: Our first try produced three dots on screen. We laughed, thinking it wouldn’t work. But pressing the down arrow revealed a functioning game! Creating a working MVP in under 3 minutes was incredible.
2. Great Collaboration: Claude was an excellent partner for my daughter and me. She’d play each iteration and give me feedback, while I added technical context to her requests. This experience rivaled or surpassed similar projects with ChatGPT and Gemini.
3. Versatile Help: Claude assisted with website design, secure coding validation, and plugin integration. Though challenging at times, it was ultimately successful.
4. Error Analysis: Claude’s error analysis was top-notch. The ability to use images to describe problems saved us lots of typing.
Challenges We Faced:
1. Code Consistency: We often lost working code snippets while fixing other issues. It felt like two steps forward, one step back.
2. Code Integration: Sometimes, small code changes wouldn’t work until we asked for a full code dump. This ate into our time limits.
3. Continuity: When Claude hit character limits, finding where to continue was tricky. A clearer “continue needed” marker would help.
4. Graphics and Design: We learned to manage our expectations for graphics. Some Photoshop work helped create our snake pieces.
5. Spatial Challenges: Handling directional errors and maze generation proved tricky, leading to some design compromises.
Overall, it was an exciting and educational experience that showcased both the potential and current limitations of AI-assisted game development.
Are you looking for a way to advance your software engineering, data, or other technology teams with AI? Are you a senior exec who is talking a visionary game about AI’s promise to their staff—while trying to learn exactly what it can do? We can help you, reach out!
Instructions:
- Use arrow keys, WASD, or touch controls to move
- Eat red food to grow
- Collect gold coins for points
- Avoid hitting yourself
- Brown blocks reduce length and change direction
- Press SPACE or Q button for Snake Quake (destroys brown tiles)
- Note: Click or tap the game board once to focus if your screen shifts
Post your high scores and suggestions and I’ll run it past my junior PO to get it added to the backlog!
