Mithril

Inspiration

Misinformation is everywhere, whether that be in casual conversation, social media, or mainstream media. Being able to access the truth in real time is more important than ever.

Despite this, conventional objective sources such as Reuters or AP News aren’t fast enough to publish articles that contain news that occurred just hours or minutes ago. To circumvent this, we harness the power of media.

What it does

We combine real time events - sourced from social media sites such as Twitter or Reddit - with objective information sources such as AP and Reuters to detect misinformation in real time.

Using dynamic agent allocation to go after real time events - we are able to quickly parse the stream of information of the internet into tangible facts.

All this in the power of a single mobile app - hold the phone out and let it pick up the conversation - it will tell you in real time if what you’re hearing is true.

How we built it

We used bright data and fetch ai for dynamic information collection via webscrapers. We used supabase and chroma db for database storage. The LLM we used to format the prompts was Gemini and the agents used Claude.

Challenges we ran into

We had lots of trouble configuring the database to the app and connecting the agents to the database.

Accomplishments that we're proud of

The dynamic agent allocation to find real time news events has lots of potential, similar to how multiprocessing works in computing. The entire flow through the database was also really satisfying to compete.