Factify

Factify was born out of a necessity, as digital citizens, to clear the fog of misinformation that clouds the feeds we, our, and everyone’s families scroll through everyday. Every time you log onto social media, you put your trust in algorithms designed for engagement rather than truth, leaving your worldview in the hands of unverified sources and an uncountable number of other—out of your hands—biases. For the first time though, as Large Language Models and Agentic AI reach early maturation stages, we can leverage these tools not just to generate content, but to verify it, creating a unified, truthful browsing experience for everyone on the web.

The societal impetus for our idea first arose out of the staggering realization that misinformation isn't just an occasional nuisance, but a regular reality; 52% of people encounter fake news regularly, and trust in platforms like Twitter and Facebook has plummeted to 16% and 11% respectively. We saw a lot of promise in the concept of "Agentic Training," and took inspiration from the idea that we could refine models to make informed decisions, essentially teaching an AI to act as a vigilant editor sitting on your shoulder.

Where Factify comes in is the unification of content consumption and verification, changing the browsing experience from a passive one to a critically engaged one. Our end goal is to move "From Chrome to the Desktop," getting our product ready for users everywhere to feel safe in their truth again. We want to answer the question: How do we give the most people access to our tool with the least friction?

Factify starts as a Chrome Extension, running a non-obtrusive UI that integrates directly with browser information. It utilizes Gemini 3 to analyze the text on the screen, cross-referencing it against a database of known winners and fact-checking sites to determine accuracy. Even in this early stage, the consumption of news transforms. For example:

A sensationalized headline becomes a nuanced discussion where you know exactly what the context is.
Scrolling through Twitter becomes less about reactive outrage and more about understanding, as the extension flags dubious claims in real-time.
The "Don't Know" factor (currently 5%) is eliminated as the tool provides immediate sourcing. As Factify matures, and we move toward a dedicated browser environment, browsing becomes an act of clarity.

What we’ve built for today is a functional Chrome Extension demonstration that overlays directly onto social media feeds. To build our extension, we leveraged the Gemini 3 model for its reasoning capabilities, Python for our backend logic, and standard HTML/CSS for the user interface.

Our biggest challenge this hackathon was balancing the complexity of agentic accuracy with the constraints of a 24-hour build cycle. We ran into specific issues regarding:

Accessibility: Ensuring the tool was available to the most people with the least amount of user friction.
Accuracy: Guaranteeing that our sourcing existed in the same information space as the content we were verifying to avoid hallucinations.
Integration: Learning how to implement and format a Chrome Extension from scratch within the time limit.

Through this process, we learned how to refine models to control the information basis of an agent, and how to design UI principles for a tool versus a dedicated environment. We discovered that for a trust-based tool, transparency and non-obtrusiveness are paramount.

To this end, what we are most proud of is our roadmap to hit the market. We have a clear timeline: Feedback and Validation immediately, publishing the extension by 12/01/25, and launching our standalone browser by 01/20/26. Looking ahead, and beyond the horizon of this hackathon, we are confident in our plan to continue developing into an entire truth ecosystem. We intend to speak with school leaders to integrate this into educational environments and continue refining our agentic training to ensure that as the internet grows, the truth grows with it.

Slides