Over the past few months, since officially joining Tupai in May, we’ve been heads-down building with AI.

In late June, Andre spent a weekend “vibe-coded” a complete student app based on everything we had learned from testing our early prototypes in schools.

With a working prototype in hand, we went straight back into classrooms, not to show off what we had built, but to find out what we had gotten wrong.

We wanted answers to a few fundamental questions:

  • Structure: How many questions should a quick practice session have?

  • Format: What works best for bite-sized practice? Multiple Choice (MCQ), Fill-in-the-Blank (FITB), or multi-step questions?

  • Engagement: What actually motivates students to keep practising, and how often will they come back?

  • Parent insights: What information do parents genuinely want to see about their child’s learning?

When building consumer apps, you quickly learn that engagement comes from constant experimentation. One idea that particularly resonated with us was that “the workaround is often the requirement.”

In other words, the unexpected ways people use your product can tell you what the product should become.

That was exactly what we were seeing in classrooms.

Students skipped things we thought were important. They found creative uses for features we hadn’t designed for. They interacted with the product in ways we simply hadn’t anticipated.

Every one of those behaviours gave us another clue about what we should build next.

From prototype to launch

With SPM just two months away and our mobile app still waiting for Google Play approval, we decided not to wait.

On September 19, 2026, we launched the web version of Tupai with a soft launch across two of our community schools.

The response has been incredible.

Over the past three days, students have kept coming back to practise. Watching them voluntarily return to something we built only weeks earlier has been incredibly rewarding and, honestly, a little surreal.

But launching is only the beginning.

The next challenge: retention and refinement

AI has given us an incredible tailwind for building and iterating quickly. As Adam Mosseri puts it, AI can be a tailwind for building consumer products.

But being able to build something quickly doesn’t mean you’ve built something great.

The real work begins when people start using it.

Our first few days have already surfaced plenty of things we need to improve: practice glitches, tricky question logic, and an auto-grading system that is, perhaps, a little too enthusiastic.

Let’s just say Tupai has some serious tiger-mom tendencies.

We’re now tuning our marking system to make feedback fairer, more accurate, and, most importantly, more constructive for students.

And that brings us to what we’re focusing on next.

Over the past few months, we’ve been figuring out how to make learning more engaging for students.

Now we want to answer a different question:

How can Tupai create even more value for parents?

Because ultimately, we don’t just want to help children practise more maths.

We want to help parents understand where their children are struggling, where they’re improving, and how they can better support them.

That’s the next chapter for us.

Try Tupai today

If you’re a parent with a child who is struggling with maths, we’d love for you to try what we’ve built.

  1. Visit parent.tupai.ai and create an account.

  2. Link your child’s profile and start practising together.

  3. Use code TUPAIFOR2 at checkout to unlock free access to all teaching and learning materials (KSSR, KSSM, and IGCSE) on the Premium Family Duo Plan for up to two children.

And if you have ideas, suggestions, or feedback, I’d genuinely love to hear them.

Drop me an email at chun@tupai.ai.

We’re still very early. Some of the best things we build next will come from listening to the people who use Tupai.