Building Classroom AI the Hard Way: Starting With Real Student Feedback

Most AI education products look impressive in demos.

Few are shaped by the people they’re meant to help.

We’re building an early-stage classroom AI, and instead of starting with big claims, we started with something simpler a demo placed directly in front of students.

No pitch deck.

No polished promises.

Just: “Try it. Tell us what breaks.”

Why We Started With a Demo

Education is sensitive. If an AI gives unclear or repetitive explanations, students notice immediately.

So rather than guessing what learners need, we decided to test early and listen carefully.

The goal wasn’t speed or scale it was understanding.

What Students Told Us

The feedback was honest and useful:

Asking questions felt easy and natural

Some answers repeated instead of adapting

Topics weren’t always separated clearly (for example, biology vs. photosynthesis)

Explanations weren’t deep enough for research-style learning

This kind of feedback is uncomfortable but essential.

Because in education, clarity matters more than cleverness.

Why Human-Centric AI Matters

Human-centric AI isn’t about branding.

It’s about behavior.

For us, it means:

noticing confusion instead of hiding it

improving explanations before adding features

letting users guide what gets built next

If an AI can’t explain a topic clearly to a student, it doesn’t matter how advanced the model behind it is.

Still Early And That’s the Point

We’re not rushing to scale. We’re not chasing hype. We’re building slowly, one insight at a time.

This is how durable education tools are made:

test early

listen deeply

improve continuously

What Comes Next

Based on student feedback, we’re focused on:

clearer topic separation

deeper explanations for study and research

better learning flow across questions

We’ll continue sharing what we learn openly, because the future of learning AI shouldn’t be built in isolation.

This project is still early.

And that’s exactly where it should be.

Try the Demo & Share Your Feedback

We’re actively inviting students, educators, and curious builders to try the early classroom demo.

Ask it real study questions.

Push it beyond simple answers.

Notice where it helps and where it falls short.

You can explore the demo here ????????:

https://baint-aio-ps-classroom-demo.vercel.app/?

Feedback matters at this stage.

Comments, criticism, and observations directly influence what we build next.

Follow the Build in Public

We’re building this openly and sharing what we learn along the way to including mistakes, fixes, and student feedback.

Follow us for updates and new demos:

X (Twitter): https://x.com/Baintcomputer?

https://www.instagram.com/baintcomputer?

Substack: https://substack.com/@baintcomputeraiops?

Here's a first-person response you can post:

This resonates deeply with me. The instinct to build first and gather feedback later is so common in edtech, and it almost always produces something that looks good in a boardroom but falls apart in an actual classroom.

What strikes me most here is the honesty about the uncomfortable feedback — answers repeating, topics blending together, explanations not going deep enough. That's exactly the kind of signal most teams would rationalize away or save for "version two." The fact that you're treating it as the foundation is the right call.

I've seen AI tools rolled into schools with great fanfare, only to quietly disappear because nobody stopped to ask students what they actually needed. Clarity really does matter more than cleverness in a learning context. A student who walks away more confused has been failed by the tool, regardless of how sophisticated the model is underneath.

The point about building slowly and sharing openly is what I'd love to see more of in this space. Most edtech development happens behind closed doors until there's something polished to announce. By then, the core assumptions are already baked in and hard to change.

Looking forward to seeing where this goes. The willingness to stay in "early" for as long as it takes is rare, and it's probably the most important decision you've made so far.

Most schools buy an AI tool, run a training session, and wonder why nobody uses it three months later. The reason is always the same: they never asked students anything before building it around them.

When I started this, I did the uncomfortable thing first. I sat with students and listened before touching any technology.

What they told me was not what I expected. They were not excited about AI doing their work. They were suspicious of it. They did not trust the outputs, felt like using it was cheating even when it was not, and several said they felt more confused after trying it than before.

That feedback rewired my entire approach. Instead of building something that gave answers, I built something that asked better questions back. The AI we use does not hand students conclusions. It asks them to explain their thinking and flags when their reasoning does not hold up. That came entirely from what students said they needed.

The second thing that came up was fairness. Students whose first language was not English said the tools felt built for someone else. That kind of feedback never shows up in a product demo. You only hear it when you actually ask.

The technical side of building classroom AI is the easy part. The hard part is creating enough trust that students tell you what is genuinely not working.

One piece of advice if you are starting this: run your feedback session before you sign any contract or write any code. Sit with students, show them nothing, and just ask where they get stuck.

The answers will change everything you thought you were going to build.