How the maths is built, and how we check it.

What Kenziq covers

Mathematics, from Nursery to Class 10, as one continuous graph rather than ten separate years. Concepts are broken into skills, each skill records what it needs to have been learned first, and the graph is built against the CBSE Mathematics syllabus for 2026–27.

We report syllabus coverage as several separate verdicts rather than one flattering percentage, because “covered” can mean a skill exists in the graph, that questions exist for it, that those questions have passed an independent check, or that a child could be assessed on it — and those are four different claims. We do not claim complete coverage of any board.

Where a question comes from

Questions are not written one at a time and they are not generated by a language model at the moment a child needs one. Each is produced by a generator — a small program that knows how to make one kind of question — driven by a seeded random number generator. The same seed produces the same question, every time, on every machine. That is what makes a bank of this size reviewable at all: a question that was checked stays checked.

The arithmetic underneath is exact. Where most software would evaluate a fraction as a decimal and accept a tiny rounding error, Kenziq keeps it as a ratio of whole numbers, so a third is a third and never 0.333…. Geometric figures are constrained so that the answer is a number a child can actually write down.

The check we care most about

It is not enough for a question to be correct mathematics with a correct answer. It has to assess the skill it is filed under. In August 2026 an audit of our own bank found a question filed under “prove that root three is irrational” that actually asked the child to find a highest common factor. Both are valid mathematics. Only one of them is that skill — and because the system trusts the label, the result was a confident estimate of a proof the child had never seen.

So every generator now declares what kind of work its questions elicit, every senior skill declares what kind of work it demands, and an automated gate quarantines the pairings that do not match. A second, independent validator re-audits the compiled bank against the syllabus. Questions that fail are held back from assessment.

We are telling you this because it is the failure most likely to recur and the one hardest to see from outside.

Practice is not proof

Everyday practice moves a working estimate of what a child can do. It does not, on its own, support a statement to a parent or a school that a child is proficient in something.

Those two things are kept apart in the system, not merely in the wording. Assessment evidence is recorded separately from practice, draws only on questions that have passed the check above, and is never pooled with practice to flatter a number. It is why you will not see a "100% mastered" badge in Kenziq.

What is in the bank today

These are counts measured from the built content, not estimates, and they are a snapshot rather than a live reading. The date they were measured is shown with them.

The unmapped figure is in that list on purpose. A distractor mapped to a specific misconception is one the system can respond to precisely; an unmapped one is still a reasonable wrong answer, but it teaches us less. We would rather publish both numbers than the flattering one.

What we do not claim

• We do not claim a child will improve by any amount. No learning-gain study has been run.
• We do not claim complete coverage of the CBSE syllabus or of any state board.
• We do not claim to detect every misconception. Many wrong answers are simply wrong answers.
• We do not claim endorsement by any teacher, school, examination board or authority.
• We do not claim the app is finished. It is in private pilot and under active development.

If an independent evaluation would be useful to you, write to us and say what you would want measured. We would rather be studied than described.

Privacy, for a school

A child's learning record carries no name, no school and no email — it is stored against a random identifier. An adult verifies a mobile number and confirms guardianship before any learning data is collected. There are no advertising kits in the app and no behavioural tracking. A parent can delete a child's profile.

If you are considering Kenziq for a group of children, the question of who holds what, and on whose consent, is one to settle before anybody starts. Write to us and we will go through it with you. The app's full privacy policy is being prepared with professional advice and will be published before release; the privacy notice for this website covers what this site itself collects.