Pundit Protocol Group 7

Frontside program

Inspiration

In each of our own lives, we've watched our own parents and the elderly in our family get trapped in cycles of polarized geopolitical views. In 2026, the greatest threat to a healthy democracy isn't a lack of information, it's a lack of perspective. That's why we decided to build Pundit Protocol.

Seeing family members get misled by single-source media outlets like Fox News or other echo chambers on social media sparked a realization: we still need to think critically even when doing something as simple as ingesting information. Fact checking was never enough. We needed to build an automated debate chat that forces multiple viewpoints into the same room to debate the truth in real-time so that we can jog the critical analysis of our user.

What it does

Pundit Protocol is an N:1 multi-agent orchestration engine. It takes a single news topic and spawns a swarm of specialized AI agents representing the entire political and professional spectrum (from articles like The Wall Street Journal and Reuters to Jacobin and Wired).

These agents perform a live, competitive debate over a WebSocket stream. The system then synthesizes their conflicting arguments into a single, nuanced truth protocol summary, making complex news digestible and safe for children, the elderly, and anyone else who may be under-informed or vulnerable.

How we built it

We leveraged Next.js 14 and Tailwind CSS to build a clean front-facing UI for our backend engine. We utilized Frame Motion for high-fidelity UI transitions and built a custom Agent Telemtry Terminal to help the user visualize the backend's workflow. We also initialized the backend software wrote its services in FastAPI, leveraging the Gemini LLM to get our desired responses for the Chaos feature.

Challenges we ran into

A challenge we ran into was that our backend processing was slow due to the amount of calls we have to do so we were able to solve this by syncing our frontend by stalling for our backend processes with a telemetry display. The toughest challenge in setting up the backend was setting up and integrating multi-agent workflows asynchronously.

Accomplishments that we're proud of

An accomplishment we're proud of is successfully moving our ideas past that of a basic GPT wrapper while still maintaining the sleek look of a modern LLM. In integrating the backend, we're really proud of our success in integrating the Gemini API for our chaos feature, as this was a successful pivot from our original plan of acquiring a Reddit API key.

What we learned

We learned a lot about how to build various visual functions and details in Next.js, some of which include: tinting the screen, smooth collapse and expand animations, and text bubbles. This goes without saying but something we're really proud of learning is pushing the limitations of integrating AI usage and learning all of the different ways that AI can work in our software.

What's next for Pundit Protocol

We really envision Pundit Protocol going a long way in serving as an accurate and high-fidelity news provider for our consumers. Pundit Protocol will adapt a modular take in the future so that our AI agents won't need to be hardcoded, but can instead be dynamically integrated as necessary with a wider variety of eligible news outlets. By doing this, we also aim to enable our users to employ specific expert agents to further debate on their industry perspective. Of course, we also aim to deploy a mobile app so that we can function as a first defense for consumers.