Saar Byrne

Maths AI Helper

AI Maths Helper

—

by

Saar
in

GitHub: [coming soon]
AI Co‑pilot: ChatGPT

Why We Built It

A friend of mine teaches maths for the Leaving Cert in Ireland. We were chatting about how students often revise on their own, stuck on a topic, without a way to get help beyond the textbook or YouTube. I had some time and curiosity, and he had the teaching experience. So we decided to try building something small.

The idea was simple:

A web app with structured maths lessons is an going project. This is still in Beta and being testing and used by a few students. What this post is about is the AI helper so that students could ask, “Can you explain this again?”

What We Made

We built a very early version that:

  • Lists maths topics for 4th, 5th, and 6th year (Leaving Cert).
  • Lets users read through basic lesson content.
  • Includes a little AI chat box where students can ask a question and get a step-by-step explanation using GPT-4.

We kept it really minimal—just enough to test the idea and get some feedback.

How We Built It

Frontend:

  • React Native with Tailwind (we’re still figuring this part out).

AI Integration:

  • OpenAI’s GPT-4 API.
  • The prompt is basic for now, but we tell the model to respond like a maths teacher explaining to a student in Ireland.

Backend:

  • Just Supabase for now—handling auth and user data.

Process:
I used ChatGPT throughout to help with:

  • API integration
  • Tailwind layouts
  • Writing prompts for the AI
  • Debugging issues with our first few attempts

I also leaned on guides and tutorials from OpenAI’s docs, Supabase examples, and a few StackOverflow posts.

Problems We Ran Into

  • Prompting was harder than I expected. At first the AI gave too much or too little info. I spent a while tweaking the instructions.
  • Designing for students felt risky. We didn’t want to overwhelm them—or mislead them if the AI got something wrong.
  • We didn’t know if anyone would even use it. This was a big one. We didn’t build analytics yet, so right now it’s mostly gut feeling and peer feedback.
  • Cost worries. The OpenAI API is powerful but can get expensive. We’ll need a smarter way to manage usage if more students start using it.

What I Learned

  • It’s okay to launch something incomplete. I usually polish forever. This time, we got the basics working and just tried it.
  • AI can help you code even if you’re not an expert. I didn’t know how to hook up the API when we started. Now I do.
  • Building with a teacher is underrated. The clarity of “this is what students actually need” helped focus the whole thing.
  • Small tools matter. Even a basic AI chat window can shift how someone learns.

What’s Next

  • Add more topics
  • Clean up the design
  • Let teachers upload their own questions or lessons
  • Possibly open it up for more students in the next term