How I think about AI in education

I built Bridging, a free AI-supported platform for Singapore O-Level Elementary Mathematics. I am also a student, which means I have used AI the lazy way and the useful way, sometimes in the same evening. This is the line I am trying to hold.

AI should shorten the distance between “I am stuck” and “I see the next step.” It should leave the step itself with the student. In math, the step is the work. A model that prints a full solution, and a student who pastes it, have both wasted the question. The design problem I care about is how to hint, check, and explain without becoming an answer key with a chat box on top.

Small things students can actually touch

Bridging’s public updates have been small and concrete. A launch. Then 40 users and more than 150 questions. Then mascots, Brix, Leslie, and Ruby, because a revision tool that feels cold gets closed. Then better math syntax, so equations, fractions, powers, and roots can look like math. Then 100 signups. None of that is a finished company. It is a sequence of things a secondary student can try.

Three constraints I do not want to drop

The syllabus comes first. O-Level E-Math is a specific exam in a specific country. A generic chatbot that “helps with homework” drifts. Bridging has to stay close to the questions Singapore secondary students are actually practising.

Cost is a product decision. If the student who needs another explanation is the one who cannot pay for another tuition add-on, a paywall on the hint is the product failing. Free is the point.

The model will be wrong sometimes. The interface has to make that easy to notice, and the student has to stay the person doing the reasoning. I would rather a tool that asks you to try again than one that performs intelligence on your behalf.

I am 17. I do not have this solved. I have a public build at bridging.lol, and a bias toward tools that respect the student. That is the standard I want AI in education to meet.