-
logo
Inspiration
As we approach graduation and the reality of leaving university sets in, we realized how much harder it is to keep in touch with everyone once the convenience of campus life is gone.
Existing social media platforms like Instagram often feel performative; it's used to "look cool" in front of strangers rather than connect with close friends. There wasn’t a space to be truly authentic. We realized that the best way to stay connected isn’t just through passive scrolling, but through shared goals and active experiences. We wanted to build a platform that bridges that gap, creating "productive intimacy" for friend groups drifting apart. That’s why we built Unitas.
What it does
Unitas is a social accountability platform designed to keep friend groups connected through three core pillars:
The Daily Challenge (Snap-Verify): Every day, the group receives a unique, "scavenger hunt" style prompt (e.g., "Find a street sign," "Touch grass," "Drink water"). Users must snap a photo to verify they did it. We use Gemini Vision to automatically analyze the image and confirm the object is present before awarding a streak.
Long-Term Goals: Users set personal goals (e.g., "Run a Marathon"), and our AI breaks them down into manageable milestones. Friends can track each other's progress on a consistency graph.
The Daily Spotlight: Instead of a noisy feed where everyone posts at once, one user is randomly selected each day to be the "Main Character" and vlog their day. This keeps the feed focused and meaningful.
Monthly Retro: At the end of the month, the app unlocks a "Reddit-style" reflection thread where friends answer deep questions about their month.
How we built it
We built Unitas with speed and user experience in mind:
Frontend: We used React Native and Tailwind CSS to build a mobile-first, responsive interface that feels like a native app.
AI Engine: We integrated the Gemini API to help break down goals into more achievable milestones. We implemented a custom ML pipeline for the "Snap-Verify" feature to recognize objects in user photos and its text capabilities to decompose user goals into milestones.
Infrastructure: The entire backend and media storage are hosted on Vultr Cloud Compute, ensuring low latency for image uploads and API requests.
Design: We utilized Lucide React for a clean, modern icon set and focused on a "dark mode" aesthetic.
Challenges we ran into State Management: Building a social app with feeds, user profiles, and dynamic challenges in a single weekend was complex. Managing the "who has posted" logic without a massive database backend required some creative state management solutions.
Defining "Authentic": It was hard to strike the balance between "productive" and "fun." We initially thought about making it strictly for work/study, but realized that was too boring. finding the "scavenger hunt" vibe took several design iterations.
Accomplishments that we're proud of Working AI Verification: Seeing the app actually recognize that we took a picture of a "red cup" or "the sky" was a magic moment. It turns a static photo feed into an interactive game.
The "Spotlight" Logic: We are proud of the "Daily Spotlight" algorithm. It solves the issue of social media fatigue by only asking one person to vlog a day, making that person feel special and keeping the content fresh.
Shipping a Clean UI: Despite the time crunch, we managed to build a polished, cohesive UI that looks ready for the App Store.
What we learned Cloud Deployment: We gained hands-on experience deploying apps to Vultr, moving away from standard Vercel deployments to manage our own compute instances.
Social Dynamics: We learned that social apps need constraints (like unlocking the feed only after you post) to drive engagement.
What's next for Unitas Notifications: Implementing real push notifications so you never miss your "Spotlight" day.
Hard Mode: We discussed an "AI Roast" feature where if you miss a streak, the AI generates a playful roast of your profile to shame you back into productivity.
Spotify Integration: We want to add a "Jam" feature to the Monthly Retro to show what the friend group listened to most that month.
Log in or sign up for Devpost to join the conversation.