Podlet

Statsig Data Demo
Home Screen
Profile Screen

Inspiration

Podlet was inspired by the spirited lunch-table discussions and Socratic seminars our team remembered from grade school. Most of the time, these were the moments when discussion pushed us to think deeper, question assumptions, and defend ideas and opinions against our best friends. We wanted to bring that same curiosity and intellectual energy into the digital age. In a time when information spreads faster than understanding, Podlet re-imagines those formative debates as an interactive, AI-driven experience that helps users engage critically with current events and other important news. It combines nostalgia for that fun, spirited conversation with a mission to rebuild thoughtful discourse through accessible technology.

What it does

Podlet transforms the day’s news into a dynamic, conversational experience. Each morning, the platform curates top articles personalized to the user and generates a podcast that summarizes them aloud. Either after or during listening, users can engage in an AI-moderated debate that challenges their understanding of current events and encourages them to back their claims with evidence. The system analyzes each response in real time, evaluating factual grounding, argument strength, and rhetorical clarity, culminating in a personalized debate score, where 100 points translates into 10 credits, which translates into $1 that can eventually be cashed out via Stripe's API after proper authentication to prevent fraud. Essentially, a debate point = 1 cent. The result is an interactive space that blends the rigor of a Socratic seminar with the accessibility of a modern news feed, encouraging critical thinking and civic literacy through conversation. This space utilizes Statsig API to send various relevant metrics that can be explored to model data helping improve the debating agents according to the user's preferences.

How we built it

Podlet's frontend is built in React and TypeScript and provides a responsive interface for browsing daily articles and participating in live debates. We integrated Firebase as our backend for real-time data storage and content retrieval, allowing the platform to dynamically load daily news from NewsAPI. ElevenLabs powers the podcast narration, converting curated summaries into natural-sounding, expressive, audio, while Gemini’s multimodal API enables the live debate system. Gemini does the processing of voice input, generating of spoken responses, and evaluating user arguments on clarity, accuracy, and critical depth. Once they have enough, the user can redeem their credits which get converted into money via Stripe's API. The Statsig API sends events from data taken during the debate hosting to the console, graphing dynamic data including various relevant metrics which, in turn, help improve the models' efficacy. Throughout development, we focused on ensuring smooth synchronization between audio playback, real-time evaluation, and database updates, resulting in an interactive and technically cohesive user experience.

Challenges we ran into

Building Podlet required bridging several complex technologies in real time that each had their own learning curves. One major challenge was integrating the Elevenlabs API with Firebase and the Gemini API to allow the conversational agent to access full news articles dynamically and accurately grade debates while maintaining a clean UI experience. Ensuring this data flow remained secure, efficient, and contextually accurate demanded careful engineering. We also faced difficulties synchronizing live audio input, output, and evaluation through Gemini’s multimodal API while maintaining low latency for a seamless debate experience. Implementing ElevenLabs audio generation alongside real-time speech processing introduced further timing and performance constraints. On the frontend, we had to optimize rendering to handle rapid state changes between debate sessions, transcript updates, and scoring feedback.

Accomplishments that we're proud of

We’re proud that Podlet became more than just a prototype and blends AI, media, and interaction into one seamless experience and the fact we were able to gamify and monetize our product. Our biggest achievement was successfully connecting a live conversational agent to real news data using Firebase and 3 separate APIs, allowing the AI to debate from up-to-date sources and be as realistic as possible. That integration of ElevenLabs narration with Gemini’s multimodal reasoning pipeline to create a podcast that transitions naturally into interactive debate is the crux of our vision. Beyond the technical milestones, we’re proud of the thoughtfulness behind the project: building a platform that encourages critical thinking and media literacy at a time when those skills are more important than ever. Seeing Podlet move from an idea about “lunch-table arguments” to a functioning, intelligent discussion partner was deeply rewarding for our entire team. It feels like we can go back to those fun debates we had, learning and growing from each other.

What we learned

Through building Podlet, we learned the importance of designing AI systems that simultaneously engage and inform. Integrating multiple APIs and data pipelines taught us how critical structured context and modular architecture are for creating intelligent, reliable behavior in real time. We gained experience coordinating cross-platform tools like Firebase, Gemini, and ElevenLabs to work together seamlessly, and learned how to manage latency, synchronization, and memory boundaries between live audio and streaming AI models. Beyond the technical lessons, we learned how to balance innovation with intention. We learned how to use AI as a catalyst for thoughtful discussion and critical thinking, not only in coding but also in our project too! The process reminded us that meaningful technology isn’t only about what’s possible, but about what encourages people to think deeper and connect over ideas.

What's next for Podlet

Our next goal is to expand Podlet with robust gamification, perhaps delving into the PVP arena. We plan to introduce user profiles, topic recommendations, and a collaborative debate mode that pairs users with others holding diverse perspectives. On the technical side, we aim to refine our debate scoring algorithm using machine learning models trained on rhetorical structure and argument quality datasets, utilizing intense Statsig-powered metrics in the process. We’re also exploring integration with additional content sources and podcast distribution platforms so Podlet can evolve from a daily news companion into a broader hub for critical discussion and civic engagement. Ultimately, we envision Podlet as a space where technology amplifies dialogue, helping people think together rather than apart.