Inspiration
Like many of us, I’ve spent countless hours doom-scrolling through news apps and social media feeds. Often, I walk away knowing only one side of the story — either the left-wing or the right-wing narrative — without truly understanding the bigger picture. This pattern not only wastes time but narrows my perspective on important issues.
That’s where the idea for DebateCast came from. I wanted to build something that would make engaging with current events more meaningful — something that would encourage thinking from both sides and spark critical conversations, even if it’s with an AI.
Using Perplexity’s Sonar API, a research-focused LLM, we could move beyond surface-level headlines to generate well-informed, nuanced points on trending topics. Whether it’s war, AI domination, or fashion trends — DebateCast helps users explore both perspectives and even lets them jump into the debate.
What it does
DebateCast is a web app that helps users understand trending topics by generating balanced, well-researched debates using AI. Users can:
Browse trending topics by category (e.g., finance, tech, world).
Choose a debate style (balanced, aggressive, funny, kid-friendly).
View an AI-generated debate with perspectives from both sides.
Select a side (Pro/Con) and engage in a live back-and-forth debate with the AI.
It’s like ChatGPT meets a debate club — helping users move beyond doomscrolling and into meaningful, critical thinking.
How we built it
Backend: Built with FastAPI in Python.
Integrates with Perplexity’s Sonar API to generate research-informed debates and trending topics.
Provides API routes for generating debates, fetching trending news, listing styles, and AI conversation for Pro/Con sides.
Frontend: Built using React.js.
We initially tried Streamlit for rapid prototyping but found it too limited for customization and interactivity.
Switched to React for better control, modern UI, and real-time debate interaction.
AI Integration: The Perplexity Sonar model served as the brain behind the debate generation and argument handling. It helped ensure high-quality, well-researched content.
Challenges we ran into
Frontend setup pain: Tailwind and Vite caused several issues on Windows, especially around module resolution and global installs. We had to rebuild our frontend environment more than once.
Prompt tuning: Crafting prompts that consistently generated high-quality debates required a lot of trial and error.
Interactivity: Simulating a real debate between the user and AI (with pro/con stance) was harder than expected, especially keeping responses consistent and coherent.
Accomplishments that we're proud of
Got a full-stack, AI-powered app working end-to-end during a hackathon.
Created a debate simulation where users can actually argue against the AI.
Built a clean, React-based frontend from scratch after pivoting away from Streamlit.
Used LLM prompting strategically to get nuanced, high-quality content — not just generic AI text.
What we learned
How to turn frustration into progress: We hit walls with frontend setup, but pushed through and shipped.
Learned the art of prompt engineering to get AI responses that sound human, are structured, and insightful.
Understood the value of well-researched AI like Sonar over generative fluff — especially for sensitive topics like war, politics, and AI ethics.
That AI is most powerful when paired with critical user input — not as a replacement for thought, but a prompt for it.
What's next for Debate Cast
Voice mode: Let users speak their arguments and hear AI responses back.
📱 Mobile support: Improve mobile UI for on-the-go debates.
🧠 Bias detector: Add an optional feature to flag potentially biased or one-sided AI answers.
🗳️ User voting: Let users vote on who won the debate — the AI or themselves.
🕸️ Topic mining: Automatically detect controversial or trending topics from news sources.
DebateCast is just getting started — and we hope it makes people think more, scroll less, and understand better.
Log in or sign up for Devpost to join the conversation.