Inspiration
When writing academic papers or preparing arguments, it's easy to get trapped in your own perspective — missing logical gaps or weak points that others might immediately spot. We wanted a tool that could automatically surface those blind spots by having two AI models debate your thesis from opposing sides, helping you stress-test your ideas before anyone else does.
What it does
AI Debate Platform lets you input any argument, thesis, or position statement, then watches as two AI agents — a Proponent and an Opponent — engage in structured multi-round debates across multiple trials. After all rounds complete, the platform produces a summary of the weakest points in your argument, the angles most vulnerable to challenge, and suggestions for strengthening your position. Think of it as a sparring partner for your ideas.
How we built it
We built the frontend as a single-page React application styled with Tailwind CSS, featuring a real-time chat-bubble interface that visually distinguishes the two AI sides (blue for Proponent, red for Opponent). On the backend, we used LangChain to orchestrate multi-turn conversational flows and connected it to the OpenAI API to power both AI agents. The debate logic handles configurable round counts and trial counts, with a final analysis pass that synthesizes all debate rounds into a structured weakness report.
Challenges we ran into
Team collaboration turned out to be one of our biggest friction points. With multiple people working across the frontend and backend simultaneously, we ran into Git merge conflicts repeatedly — especially around shared state logic and component files. Resolving these took a significant amount of time and taught us hard lessons about branching strategies and communication before pushing changes.
Accomplishments that we're proud of
We're proud that the platform turned out to have far broader applications than we initially imagined. Beyond academic research, it works well for policy analysis, business proposal review, debate practice, and really any scenario where you need to anticipate counterarguments. Seeing the AI surface genuine weaknesses in carefully constructed arguments — ones we hadn't thought of ourselves — was a genuinely exciting moment.
What we learned
Building this project taught us how to use AI not just as a generator of content, but as a structured reasoning tool that produces genuinely valuable analytical output. In an era where AI is everywhere, the real skill is in designing the right prompts, flows, and evaluation criteria so the output is meaningful — not just plausible-sounding text.
What's next for AI Debate Platform
We plan to keep iterating with AI at the core — making the platform smarter, faster, and more accessible. Future directions include supporting user-defined AI personas and debate styles, PDF/CSV export of weakness reports, multilingual support, and expanding into use cases like legal argument review and product strategy evaluation. The goal is to build something that genuinely helps people think more rigorously and make better decisions.
Built With
- docker
- fastapi
- github
- javascript
- langchain
- openai
- postgresql
- python
- vue
Log in or sign up for Devpost to join the conversation.