The tutor listens first. Then it asks.
Enumora is not a chatbot a child types into. It is a spoken, one-to-one tutor: the child works through a lesson aligned to the school syllabus while the tutor listens to their reasoning, spots the specific misconception, and answers it with a question. The child does the thinking. We guide children through concepts — we never hand them answers.
- Voice-first. The tutor speaks; the child answers aloud. Speech is recognised on the phone, never uploaded.
- Curriculum-true. Lessons follow the school’s board and chapter, starting with NCERT-aligned Class 7.
- High-dosage pattern. Short, frequent, one-to-one sessions — the tutoring pattern education research links to real progress.
- Offline. Everything above happens with no internet at all.
Anatomy of a lesson
Eight moves between a misconception and mastery.
This is the photosynthesis session from the product, word for word. The Pedagogy view is on, so you can see what the lesson engine is doing at each turn. Every move below is enforced by the engine, not left to the AI’s mood.
Recall hook
The lesson opens with the child’s own words from yesterday — then a prediction question.
“If I keep a plant in a dark room, with the best soil — will it stay healthy?”Misconception classified
The answer is matched against the chapter’s misconception bank. Here: “soil is food”, the most common error in the chapter. The tutor never says “wrong”.
Counter-evidence question
No correction. A puzzle that makes the child confront their own model: why would leaves turn toward the light?
Scaffold ladder
Support rises step by step, aimed at the exact mistake, and never drops while the child is still stuck. Concept first, vocabulary last.
The child names it
“Photosynthesis” arrives from the child, after the idea is owned — the self-explanation effect.
Retrieval check
The opening question, asked again. Only an answer that carries the mechanism counts — not a memorised word.
Stars for mastery
The tutor awards exactly the stars it says aloud, and only for secured understanding. A session can also end kindly with none.
Forward hook
Tomorrow’s thread — stomata, the leaf’s tiny doors — so the child returns to a story, not a chore.
The rule we enforce in code
“Correct the child directly” is treated as a bug.
Every lesson is a script written by our pedagogy team and checked by a linter before it ships. The AI classifies the child’s answer, asks follow-ups within a phase and speaks naturally in the child’s language — but it never freelances the pedagogy.
- Answer-leak refusal. Before the child hears any tutor turn, the engine checks it against the topic’s key mechanism in the spoken language. A turn that would hand over the answer is withheld.
- Honest stars. The number of stars spoken always equals the number awarded. Every ending of every script is tested for it.
- Read by humans. Transcripts of every lesson path, in every language, are produced on each build — and read, preferably aloud, by people.
- Safe pauses. If a child says something that suggests distress, the tutor stops teaching and shows one calm card asking them to find their teacher. No sirens, no red.
The child’s day
Goals the child chooses. Sessions the teacher sets. Never one muddled list.
The Today screen holds two different things. Goals are self-paced and can be done in any order — start anywhere, and the tutor keeps every subject’s thread. Sessions are appointments the teacher has set, with outcomes she will see.
- Every card ends in exactly one action. Continue the lesson, start the quiz, preview the session, see the outcome. A dead-end card is a bug.
- Three doors, one tutor. Self-paced lessons, teacher-scheduled sessions and practice all reach the same tutoring surface.
- Sign-in a child can do alone. Tap your picture, then a 4-digit secret or a three-picture “story”. Nothing typed, nothing spelled.
Mastery Stars
The only reward is understanding.
Most learning apps pay children for time: streak bonuses, combos, login rewards. That teaches children to stay, not to learn. Enumora’s star economy has five laws, and the product is built so it cannot break them.
Mastery only
Stars come from a quiz cleared with reasoning, a topic secured in dialogue, or a credential signed. Never from minutes, streaks, combos or logins.
Revision shows ⭐ +0
Spaced revision brings back secured topics just before forgetting sets in. It earns nothing — and the card says so honestly.
Capped by the syllabus
The total a child can earn is bound to what has been taught so far. The counter cannot inflate, so a star always means something.
Streaks pause, never shame
A missed day pauses the streak. There is no guilt copy anywhere, and the words “wrong”, “failed” and “locked” never appear on a child’s screen.
Every number explains itself
Tap any figure and it tells you what it means — “mastery of what has been taught so far, not exam marks”.
When learning happens
Twelve occasions, three people who can start them.
If a child’s motivation dips, the teacher’s scheduling sustains the habit. If the teacher is stretched, self-paced goals and gentle system suggestions carry it. No single person has to hold the whole thing up.
| Context | Teacher-set | Child-started | System-suggested |
|---|---|---|---|
| Class or study period | Scheduled one-to-one session, three times a week | — | — |
| Free period | Re-teach set for a diagnosed misconception | Continue a lesson, practise ahead | Topic quiz toward a credential |
| Evening at home | — | Ask a doubt, photograph homework | Revision, honestly ⭐ +0 |
| Weekend | — | Reading fluency, a lesson | — |
| Exam season | Drill schedule | Revision | Retrieval checks |
| Holidays | — | Practise ahead — no waiting for chapters | Keep the thread warm |
| Back from an absence | Session re-slotted | — | Catch-up: “the tutor kept your thread” |
We create occasions, but we measure mastery per occasion — never minutes.
Why offline
The internet ends where most childhoods begin.
Enumora’s tutor is a compact AI model that lives on the phone. The teaching happens entirely offline — in the child’s evening, the village night, the power-cut hour. The cloud enriches; it is never the path between a child and their lesson.
Teaching happens here, offline
- Socratic dialogue in Hindi or English
- Quizzes, graphs and diagrams, generated on the phone
- Speech recognition and the tutor’s voice
- Credentials signed the moment mastery is verified
- Conversations stay here. Always.
What the school sees
- Which topics the class has secured
- Where misconceptions cluster
- Credentials earned, verifiable in a second
- Never a child’s words
What we won’t claim. No AI on earth generates video on an ordinary phone, and we will never pretend otherwise. Lesson videos and illustrations are delivered in content packs whenever the phone next meets Wi-Fi. Offline means offline at the point of use — which is where learning actually happens.
Languages
A language ships only when a child can be taught in it — not merely addressed.
The app speaks nine languages today, with full right-to-left support for Arabic. The tutor teaches in Hindi and English first, because teaching Science in a language takes far more than translating its buttons. More teaching languages join one at a time, each with its own misconception bank and a tutor voice children trust.
The app, today
English · हिन्दी · ଓଡ଼ିଆ · தமிழ் · తెలుగు · বাংলা · मराठी · ગુજરાતી · العربية
The tutor, at launch
Hindi and English, including the natural Hinglish of a real classroom. Hindi-first authoring; English co-authored.
Subjects in the founding cohort
Classes 6–8: Science, Mathematics, Hindi and English, starting from a full-year Class 7 spine.
Hear it in your own classroom.
We’ll run a lesson with your Class 7 children — with the Wi-Fi off. If it doesn’t teach, you’ll know in thirty minutes.