An AI tutor that works with a teacher, not instead of one
We are building an AI tutor for Class 1 to 10 - a study partner your child can use between classes, on the practice that needs repeating. It does not exist yet. There is nothing to sign into and nothing to try, and we would rather say that plainly than dress up a waiting list as a product.
What we can tell you is the shape of it, and what we have already decided it will refuse to do. Those decisions matter more than the launch date.
Available right now instead
Four rules we are building it around
Most of what makes an AI tutor good or harmful is decided before a line of it is written. These are ours, published now so you can hold us to them.
It will not do the homework
A tool that hands over answers teaches a child to ask for answers. Ours will ask questions back, give one hint at a time, and stop short of the final step so the child takes it themselves.
It works for the teacher, not instead of them
The teacher sets what to practise and sees where a child got stuck. The AI covers the repetition between classes - the twentieth practice sum, the spelling drill at 9pm - not the teaching.
It follows your child's actual syllabus
Class, board and chapter, matched to the book on your table. Grounded in the real NCERT progression rather than generic content scraped from anywhere.
Built for children, with the caution that deserves
No advertising, no engagement tricks, no data sold. A parent will be able to see every conversation their child has with it.
What AI is genuinely good at, and where it fails a child
This applies to whatever your child is already using, not only to what we build. The split is fairly consistent, and it is not the split most people expect - the failures are not about knowledge.
| Task | How well it goes | Why |
|---|---|---|
| Endless practice at a fixed level | Very well | There is no judgement involved, and generating the two-hundredth question is as cheap as the first. |
| Explaining a concept a second way | Well | Rephrasing is what these models are best at, and a child who did not follow the first explanation often follows the third. |
| Marking a closed answer | Well | Right or wrong is checkable. Whether the working deserves partial credit is much less so. |
| Working out why a child got it wrong | Poorly | A wrong division answer can be division, tables, place value or a misread question. They look identical on the page and need opposite responses. |
| Noticing what a child is quietly avoiding | Not at all | It only sees what is typed. The child who stops asking, or picks the easy exercise every time, is invisible to it and obvious to a teacher. |
| Being wrong convincingly | A real risk | A child cannot tell a confident wrong explanation from a right one, and one learned early is harder to undo than no explanation at all. |
Why we are being slow about this
Plenty of products already promise an AI tutor. Some are genuinely good. Many are a chat box that answers the question, which is the easiest thing to build and close to the worst thing for a child - it teaches them that the way out of being stuck is to ask a machine.
There is also a real risk we take seriously: an AI that is confidently wrong. A child cannot tell the difference, and a wrong explanation learned early is harder to undo than no explanation at all. Grounding it in the actual syllabus, and having a teacher check the difficult ground, is slower than shipping a chat box. We think it is the only version worth building.
And it is going in front of children. India’s DPDP Act 2023 sets real obligations around a child’s data, including verifiable parental consent. We would rather clear that properly than launch and apologise.
What will not change
- A real teacher stays at the centre of how we teach
- Nothing you already pay for moves behind a new charge
- No advertising to children, ever
- Parents can read every conversation their child has
- We will say when it is ready - not before
14+ years of classroom teaching sits behind these decisions. It is why the teacher stays in the middle of them.
Join the waitlist
Tell us your child’s class and we will message you when there is something real to try - not before. Waitlist members get it free throughout testing, and we will ask what actually helps rather than guessing.
You can also just book a free assessment and meet a teacher this week, which is available today.
Questions about the AI tutor
Can I use this today?
No. It is in development and there is nothing to try yet. Join the waitlist and we will message you when there is. In the meantime our live classes with a real teacher are running now, and the worksheets and games are free to use.
Will this replace the teacher?
No, and we would not build it if it did. A child who is stuck is usually stuck for a reason software cannot see - they are tired, they missed a week, they are embarrassed to ask. A teacher notices that. The AI will handle repetition between classes so the teacher's time goes to the part that actually needs a person.
Will it just give my child the answers?
It is being built specifically not to. Handing over answers is the easiest thing to build and the least useful thing for a child. It will ask what they have tried, give one hint, and let them take the last step - the same thing a good teacher does.
What will it cost?
Not decided. Waitlist members will get it free throughout testing, and we will tell you the price before it ever becomes paid. We will not move anything you already pay for behind a new charge - our current fees are published openly.
What about my child's privacy?
This is the part we are slowest and most careful about. India’s DPDP Act 2023 requires verifiable parental consent before processing a child’s data, and we will not launch until that is properly handled. There will be no advertising to children, no engagement mechanics designed to keep them on it longer, and a parent will be able to read every conversation. Our current privacy policy explains what we collect today.
When will it be ready?
We are not giving a date. Software dates given early are usually wrong, and this is going in front of children, so we would rather be late than careless. The waitlist is the honest answer - you will hear from us when there is something real.
Is there a free AI tutor for kids?
Several, and the general-purpose chatbots are the ones most children actually use. They are genuinely capable and they are not built for this: they answer, which is the fastest way to teach a child that the way out of being stuck is to ask a machine. If your child is already using one, the useful intervention is not banning it but changing the instruction - ask it to give a hint and not the answer, and see whether it obeys.
Can an AI tutor help with reading?
Partly, and less than the marketing suggests. Listening to a child read aloud and catching the specific decoding error - a blend skipped, a vowel guessed from the picture - is exactly the judgement that is hardest to automate reliably, and getting it wrong reinforces the error. Reading practice, word lists and comprehension are useful uses; diagnosing why a child cannot blend is not one yet.
Is it useful for maths?
Maths is where AI is strongest and where the failure is most expensive. It is strong on generating unlimited practice at a fixed level, which is genuinely tedious to do by hand. It is weak on why a child got something wrong - a wrong answer to a division question can be a division problem, a tables problem, a place-value problem or a misread question, and those need opposite responses.
Can we use it for homeschooling?
As practice and explanation, yes. As the teacher, not yet, and being straight about that matters more here than anywhere - a homeschooling family has no second system to catch what the first one misses. The gap is not knowledge, which AI has plenty of. It is noticing what a child has quietly not understood while appearing to keep up.