Steelman

Inspiration

Our main inspiration is that we want people to see the other side of their perspective. Not to change their mind, but to expand it.

What it does

Steelman helps you fact-check and critically evaluate claims by generating the strongest possible counter-arguments using AI. Instead of attacking weak arguments, it presents the strongest version of opposing viewpoints with evidence and sources, helping you see different perspectives and make more informed decisions.

How we built it

It's a Next.js frontend with an Express.js backend, using OpenRouter API for AI-powered counter-arguments. The frontend is on Vercel, backend on Railway, with PostgreSQL for the database. Built with TypeScript throughout for type safety.

Challenges we ran into

Main challenges: AI returning malformed JSON (fixed with multi-strategy repair), React hydration errors from localStorage (fixed with proper state initialization), deployment issues on Vercel/Railway (fixed with proper configs), API migration to OpenRouter, and cache/history management. Most were solved with better error handling, type safety, and deployment configuration.

Accomplishments that we're proud of

Steelman's core features include AI-generated counter-arguments with evidence and sources, article analysis that extracts claims and generates counter-arguments, a history sidebar with search and filtering, copy functionality for claims and arguments, dark/light mode, real-time processing status updates, multi-tier caching for fast access, robust JSON parsing for handling AI responses, and a mobile-first responsive design. The most impactful are the steelman counter-argument generation with evidence and the history sidebar for quick access to past analyses.

What we learned

With the skills learned and developed, we should be able to build robust error handling, use TypeScript strictly, handle SSR state carefully, and abstract APIs for flexibility.

What's next for Steelman

Deploy to production (Vercel for frontend, Railway for backend with PostgreSQL), then add user authentication, improve mobile UX, and add analytics. Future features include model selection, export/share, batch analysis, browser extension, and an API. Priority: get it live, monitor performance, then iterate based on user feedback.