ReTrace

• 

  Logo

• 

  Main page. Used for uploading the problem.

• 

  Analysis of the problem example.

• 

  Leaderboard

Inspiration

Most AI learning tools give answers or explanations. ReasonTrace focuses on why a student got something wrong.

What it does

ReTrace analyzes a student’s solution, finds the first incorrect reasoning step, identifies the likely misconception, and generates a quick diagnostic question to verify it. It then gives a short explanation and targeted practice.

How I built it

I built ReasonTrace with Next.js, TypeScript, Tailwind CSS, Gemini API, and Vercel. Gemini returns structured reasoning data that the app turns into a visual step-by-step trace.

Challenges

The main challenge was making the AI diagnose reasoning instead of immediately giving the correct answer. I also had to distinguish misconceptions from simple calculation or careless mistakes.

What I learned

We learned that AI feedback becomes more useful when it tests its diagnosis before teaching. A focused reasoning workflow can provide more precise support than a general AI chatbot.

Built With

• api
• css
• gemini
• html
• javascript
• json
• next.js
• node.js
• react
• rest
• tailwind
• typescript
• vercel