MathMagic

Project Goal

The main goal of MathMagic is to leverage the latest AI models to solve high school and university-level math problems accurately. Historically, no AI model has been able to solve math exercises with high accuracy. However, with the recent release of OpenAI’s o1 model, I have noticed significant improvements.

I developed MathMagic, starting with a demo version created about six months prior to explore AI’s potential for solving complex math problems. At the time, the available AI models were costly and not yet ready for broad use. The project progressed significantly with the advent of more stable and affordable models, such as OpenAI’s o3-mini, which allowed me to refine the application and make it more practical.

Tech Stack Deep Dive

• Frontend: Next.js, Tailwind CSS
• Backend: Node.js, MongoDB
• AI Models: Gemini, OpenAI
• Infrastructure: Google Cloud Platform (GCP), Cloud Run

Challenges & Solutions

One of the strengths of MathMagic is its high accuracy. I tested it on various math exams, and it consistently scored above 90%. It solves simpler questions faster and with higher accuracy. However, it still struggles with complex visualizations like graphs, which can result in lower success rates. Therefore, while the results are not 100% accurate, the detailed explanations provided can guide students effectively.

Key Learnings

I learned that formatting a large document (math solution) requires a thinking model for stable and accurate solution.