This past weekend, I did what every developer dreams about but rarely executes: I went from zero to deployed on not one, but two complete web applications. Armed with Google AI Studio, a Git repository, and Vercel’s deployment pipeline, I turned coffee and curiosity into live websites pointing at my own domains.
Here’s how it went down.
The Setup: Modern Dev Stack in Minutes
The beauty of today’s development ecosystem is how quickly you can go from concept to production. My stack was deliberately simple:
- Google AI Studio for rapid prototyping and AI integration
- Git for version control (because we’re not animals)
- Vercel for deployment and hosting
- Custom domains to make it official
The entire setup—from initializing repos to configuring DNS—took less time than my morning coffee routine.
Project One: Top-Down Vector Shooter at ijduncan.com
First up was something purely for fun: a browser-based, top-down vector shooter. Think classic arcade aesthetics meets modern web technologies.
The Build
Using Google AI Studio, I scaffolded out the game logic quickly. The vector-based graphics gave it that clean, geometric feel—all clean lines and smooth movement. The physics were surprisingly satisfying to dial in, and within a few hours I had:
- Smooth player movement and controls
- Enemy spawn mechanics
- Collision detection
- A scoring system
- Vector-based particle effects for that retro-future vibe
The Deploy
I pushed from Google ai studio directly to my git and then over to Vercel:
- Import Git repository
- Configure build settings
- Deploy
- Point ijduncan.com at the deployment
15 minutes later, I had a playable game running on my own domain.
Project Two: AI Fitness Coach at jeffphillipsfitness.com
The second project was more practical but equally satisfying: a single-page website for my personal trainer friend Jeff, complete with an embedded Gemini-powered chat agent.
The Vision
Jeff needed a web presence that could actually help potential clients. Not just a static brochure site, but something interactive that could answer questions about training, nutrition, and fitness goals 24/7.
The Build
Google AI Studio’s Gemini integration made this almost trivially easy. I built out:
- A clean, single-page layout focused on Jeff’s services
- An embedded chat interface (bottom right corner, naturally)
- A Gemini-powered agent trained on fitness and training knowledge
- Prompt engineering to keep responses professional, encouraging, and on-brand
The chat agent can discuss:
- Training methodologies and program design
- Nutrition basics and macro planning
- Injury prevention and mobility work
- Scheduling and service inquiries
I also used gen AI to create many of the images the video banner. There was not a video of Jeff. That was created with image to video in Veo.
The Magic of AI Agents
What’s fascinating is how quickly you can create a specialized AI assistant. With the right system prompts, the Gemini model stays in character as a knowledgeable fitness professional, steering conversations toward health and training topics while maintaining Jeff’s approachable, no-BS coaching style.
Deploy and Point
Same smooth process:
Vercel handled the build and deployment. Just add a google API key for Gemini and that was it. Point jeffphillipsfitness.com at it, and suddenly Jeff has a live site with an AI receptionist that never sleeps.
Watching It All Come Together
There’s something deeply satisfying about the modern deployment workflow. Push to Git, watch Vercel’s build logs stream by, see the deployment go live, hit your custom domain, and there it is, your thing, live on the internet, accessible to anyone.
No server configuration. No certificate headaches. No deployment scripts to debug. Just code, commit, deploy, done.
What I Learned
1. Vercel’s Developer Experience Is Unmatched
The integration between Git and Vercel is seamless. Every push triggers a new deployment. The production deployment happens automatically on merge to main. DNS configuration is straightforward. It’s how deployment should always feel.
2. AI Studio Accelerates Prototyping
Google AI Studio let me iterate rapidly on both projects. For the shooter, I could quickly test game logic variations. For the fitness site, I could refine the chat agent’s personality and knowledge base without rebuilding infrastructure.
The Weekend Scorecard
Time invested: ~6 hours across both projects
Deployment headaches: Zero
Live sites: Two
Satisfaction level: Maximum
What’s Next
Both sites are live and functional, but they’re far from finished. The shooter needs sound effects, power-ups, and maybe a leaderboard. The fitness site could use a blog section, and deeper integration with Jeff’s actual booking calendar.
But that’s the beauty of this workflow—iteration is easy. Each improvement is just another commit away from being live.
Try Them Out
- Play the shooter: ijduncan.com
- Chat with the AI fitness coach: jeffphillipsfitness.com
And if you’re thinking about a weekend build of your own, my advice is simple: pick your stack, start typing, and let the deployment pipeline handle the rest. The modern web makes it easier than ever to go from idea to live site in a weekend.