## Inspiration
The Olympic Games often feel like a distant spectacle—extraordinary humans doing impossible things on a screen. With the Games coming to Los Angeles in 2028, we wanted to bridge the gap between the "fan" and the "athlete." We were inspired by the idea of "Vibe Coding"—using AI to transform raw athletic data into a personal, relatable, and high-energy experience. We asked: What if you could see yourself in the roster? What if you knew exactly when your "athletic twin" was competing?
## What it does
LA28 All-In is an AI-native intelligence engine designed to turn every fan into an Olympic insider.
- The Olympian Mirror: Users input their physical stats and archetypes to find their Team USA "twin."
- Team USA Selection Tracker: A live-updating roster that celebrates athletes the moment they are officially named to the LA28 team with custom badges and congratulatory messages.
- What-If Engine: A simulation tool that predicts medal probabilities and performance outcomes using complex reasoning.
- AI Commentator: A dynamic WebSocket-driven feature that provides "TV-ready," sport-specific live commentary for trending events.
- Hybrid Notifications: A multi-platform alert system (SMS, Email, X, etc.) that keeps fans locked into the competition schedule.
## How we built it
We utilized a modern, agentic stack focused on rapid iteration and intelligence:
- The Brain: Gemini 3 Pro via Google AI Studio. We leveraged its long-context window to ingest massive athlete rosters and its Search Grounding to find real-world competition dates.
- The Backbone: A FastAPI Python backend handling WebSocket streams for the live commentary and the simulation logic.
- The Logic: We used LaTeX-based formulas to calculate "Match Percentages" for the Olympian Mirror, ensuring the math behind the "vibe" was sound: $$Match \% = \left( 1 - \frac{|\text{User}{metric} - \text{Athlete}{metric}|}{\text{Athlete}_{metric}} \right) \times 100$$
- The Workflow: Developed using Antigravity IDE and Vibe Coding principles, allowing us to move from natural language prompts to functional UI components in record time.
## Challenges we ran into
The biggest hurdle was "Data Freshness." Olympic qualification cycles are fluid. Originally, our roster was static, but we realized that to be useful, it had to be dynamic. We had to pivot from hardcoded dictionaries to a model-driven approach where Gemini 3 Pro fetches the latest trials results in real-time. Additionally, synchronizing the WebSocket commentary to feel "live" without lag required careful asynchronous management in Python.
## Accomplishments that we're proud of
We are incredibly proud of the AI Commentator. It doesn't just spit out generic text; it understands the "vibe" of different sports. It switches tone between the explosive energy of 100m Track & Field and the poised technicality of Artistic Gymnastics flawlessly. We also successfully implemented a "Fallback Notification" system that ensures the UI never breaks, even if third-party API keys (like Twilio) aren't present.
## What we learned
We learned that Agentic Workflows are the future of development. By using the Agent Development Kit (ADK), we stopped writing boilerplate and started "guiding" the code. We also discovered how powerful Search Grounding is for niche topics like Olympic trial schedules—it transformed our app from a static demo into a live utility.
## What's next for LA28 All-In: Olympic Intelligence Engine
The road to 2028 is long! Our next steps include:
- Full Firebase Integration: Moving our simulated notifications into real-world Firebase Cloud Messaging (FCM).
- AR Mirror: Using MediaPipe to allow users to "pose" like their matched Olympian in augmented reality.
- Community Hub: A "Fan Zone" where users matched to the same athlete can coordinate watch parties and "Vibe Check" the latest competition results together.
Built With
- aistudio
- antigravity
- firebase
- gemini3pro
Log in or sign up for Devpost to join the conversation.