Inspiration
Honestly, growing up, I was one of those kids who fell for a lot of those crazy conspiracy theories like "Tupac never died, and he's hiding out in Cuba". I'm fortunate enough to say I outgrew those days. But, those are fairly harmless things to believe in comparison to the mis/disinformation that flourish all over the internet, especially on social media. Today, there is hardly any consensus at all and during the pandemic, seeing 800+ people had lost their lives because they read the wrong information online made me want to step up and address this problem. But, back then, I didn't know nearly what I know today and with sophisticated coding agents, building aspirational systems like what I have built are not as daunting. The goal is to tear down digital tribalism's walls and rebuild consensus in the age of misinformation and AI.
What it does
You get access to a media analyzer that can analyze videos, articles, and social media posts (from select platforms like substack and truth social). Every time you analyze content, your history is saved so you can always access it in the future. Furthermore, there are feeds of the most substantiated content that we offer for free as well (WIP, but visually there). I also laid the groundwork for gamification to be integrated at a later date and more social components like group chats with your friends and family who desire to be well-informed!
How we built it
I created a Next.js web app, tied it to the google firebase suite including auth and other GCP services to manage data like signing in through Discord, Gmail, setting up user profiles, and of course handling the media that users submit by link for analysis. I use NextJS for a simple API to act as an intermediary to route trafficking too. However, that is not my main focus in comparison to the front-end. I spent far more time on that, where most of the code resides. I also leveraged a lot of existing APIs like Exa and Firecrawl's generous free tiers to setup and experiment with web scraping (although they could not break through Twitter either). Simple cloud run functions in GCP helped keep things professional and tidy so that my codebase didn't get too overwhelming.
Challenges we ran into
I could not break through Twitter's firewall, so I am unable to scrape their content currently. That sucked some time away. Also, managing schemas for how to render data in the frontend that is produced by an API or database was not so easy. Trying to consolidate a variety of similar media analysis modes that required displaying varying data was difficult, even with AI support to devise a more monolithic schema to follow.
Accomplishments that we're proud of
I am most proud of the multimodal support that my platform has. We can analyze Youtube Videos, Tiktoks, Substack articles, Truth Social posts, and mainstream media articles. We only plan to support more media formats!
What we learned
I learned how complex circumventing firewalls, CDNs, and managing states in a full-stack application can be. I want to continue learning about good software design principles so that this project becomes a beloved platform for misinformation mitigation. I also learned a bit about how ontologies are made for AI models, which I can continue learning about after the hackathon even if I didn't get to implement a proper one.
What's next for Reef Media
Build out the enterprise dashboard component at O1 House's Invitational Hackathon in mid Apirl so that we can complete the full E2E experience and begin facilitating pilots with media platforms that are under increased regulatory pressure to crackdown on misinformation, but lack an affordable way to do it! Social Media platforms are having their big tobacco moment as we speak in front of Congress! Timing could not be better.
Log in or sign up for Devpost to join the conversation.