Every lesson
is a lab.
Teachers and AI write the curriculum together — then students learn it by running the experiment, changing one variable, and watching what breaks.
Free for teachers. No card, no ads, no student data sold.
How a lesson gets built
A teacher decides what a child should understand and how it should be taught. The AI does the volume work — drafting, varying, translating, checking. Neither half ships alone.
Sets the target
A classroom teacher names the misconception to fix — "students think heavier balls fall faster" — and the standard it maps to.
Drafts and varies
Claude writes the explanation, builds the simulation parameters, and generates 40 practice variants at six reading levels in nine languages.
Cuts and corrects
Two subject specialists rewrite what's wrong, dull, or age-inappropriate. Nothing reaches a child until a named human signs it.
Tested with kids
The lesson runs in real classrooms. Where students stall, the section gets rebuilt — usually within a week.
AI draft, lesson 6.2
"Acceleration due to gravity is a constant 9.81 m/s², independent of mass, as demonstrated by the equivalence principle."
After teacher review
"Drop a marble and a bowling ball together. They land together. Try it in the sim — change the mass slider as much as you like and watch the clock refuse to budge."
Reviewed by D. Okafor, 11 years teaching middle-school physics. Every lesson page carries the reviewer's name.
The interactive lab
These four run right here, in the page — no login, no download. It's the same engine students use, with the same controls.
Hit the target
Two different angles can land in the same place. Find both.
Set an angle and launch.
Mix to neutral
Get the beaker to pH 7. Then try it again using half as much water — does the recipe change?
Acid: hydrochloric, 0.1 mol/L.
Base: sodium hydroxide, 0.1 mol/L.
Colour follows universal indicator.
Start with 100 mL of pure water at pH 7.
Hares and lynx
Watch the two curves. Which one peaks first — and why does it have to be that one?
Green is hares. Amber is lynx.
Why the bottoms must match
Slide the pieces together. The bars only stack neatly when the two fractions are cut the same way.
Top row sets how many parts. Bottom row sets how many the whole is cut into.
1/2 + 1/3 = 3/6 + 2/6 = 5/6
What's in the library
Every lab is tagged to a learning outcome and a chapter, so a teacher can find what fits tomorrow's period in under a minute. Coverage deepens each term.
| Subject | K–2 | 3–5 | 6–8 | 9–12 | Languages |
|---|---|---|---|---|---|
| Physics | — | 34 labs | 118 labs | 196 labs | 9 |
| Chemistry | — | 21 labs | 86 labs | 174 labs | 9 |
| Biology | 18 labs | 52 labs | 94 labs | 142 labs | 9 |
| Mathematics | 46 labs | 128 labs | 161 labs | 139 labs | 11 |
| Earth & Space | 12 labs | 38 labs | 64 labs | 47 labs | 7 |
| Computational thinking | 9 labs | 31 labs | 58 labs | 72 labs | 5 |
Counts as of this term. Figures shown throughout this page are illustrative and should be replaced with your own verified data before publication.
The same idea, cut four ways
A six-year-old and a sixteen-year-old can both run the pendulum. What changes is the question we ask them afterwards.
Touch first, words later. One control, big targets, spoken instructions, no reading required to start.
Push the swing harder and count how far it goes. Sort objects that float from ones that sink.
Predict, then check. Students write down a guess before the sim runs, so being wrong becomes the interesting part.
Build a circuit that lights two bulbs. Feed a food web and watch what happens when the frogs disappear.
Change one thing at a time. Controlled variables, data tables, and a graph the student plots from their own runs.
Find the launch angle for maximum range. Titrate to neutral. Model a population crash.
Math behind the motion. The equations governing the sim are exposed, editable, and matched to exam specifications.
Derive the range formula and test it against the sim. Fit a rate law to your own titration data.
Does it actually work?
The honest answer is that simulation alone changes little. Simulation plus a teacher who asks the right question afterwards changes a lot. Here is what we measure, and what we don't claim.
On conceptual questions
Mean gain on pre- and post-tests of understanding, measured across 41 Grade 6–8 classrooms over one term, against matched control classes.
More student talk
Classroom audio sampling showed students spoke more than three times as often during lab periods than during the same chapter taught from the textbook.
Prep time per lesson
Teachers reported less time spent building worksheets and differentiated practice, and more spent planning the discussion that follows the lab.
What we don't claim
- Board exam scoresOne term is too short, and too many other things change at once. We are running a three-year study and will publish it whatever it shows.
- That it works without a teacherIn our own pilots, unsupervised lab time produced almost no measurable gain. The question asked afterwards is where the learning happens.
A district in practice
A 14-school district began with one chapter in Grade 6 science across four schools. Teachers kept their own textbooks and their own sequence; the only change was a single lab period per chapter.
After two terms, 11 of the 14 schools had adopted it across all middle-school science. The three that didn't cited device availability, which is the constraint we hear most often and the one we design around.
What each half is for
The teacher decides
- What counts as understandingEvery lab ends in a question a machine can't score by keyword matching.
- What a child of this age can handlePacing, reading level, and how much failure is productive before it's just discouraging.
- What gets cutReviewers delete roughly a third of every AI draft. That's the job working, not failing.
- The final sign-offNo lesson publishes without a named subject specialist attached to it.
The AI does the volume
- Forty versions of one worksheetSo a class of thirty gets thirty different problems at their own level.
- Translation and reading levelsNine languages, six reading levels, generated from a single reviewed source.
- Live hints during a labNudges toward the next step instead of handing over the answer.
- Spotting where a class stalledFlags the three students stuck on the same slider, before the lesson ends.
Where we're careful
- AI never talks to a child unsupervisedHints are pre-reviewed and scoped to the current lab. There's no open-ended chatbot for students.
- Student work stays with the schoolNever sold, never used to train a public model. Districts can export or delete everything, any time.
- Teachers can overrule anythingEdit a lesson, hide a lab, change a score. The classroom is not run by the software.
What a school gets
- Curriculum mappingEach lab tagged to your state or national standards, with gap reporting.
- Runs on what you ownChromebooks, shared tablets, a single projector at the front. Nothing to install.
- Onboarding in an afternoonTwo hours of teacher training, then a specialist joins your first three lessons.
Why does the Moon change shape?
A complete 40-minute period for Grade 6 — lamp-and-ball demo, then a simulator showing the Sun–Earth–Moon system from above beside the Moon as seen from your own street. Built to fix the misconception that phases are the Earth's shadow.
Getting a school running
Most schools are teaching their first lab within a fortnight. Nothing is installed, nothing is migrated, and no existing textbook is thrown away.
Scoping call
Forty minutes with your science and maths leads. We map the labs to your scheme of work and check what devices you actually have in each room.
Teacher training
One two-hour session, run in your school or online. Teachers leave with a lesson they have already taught to each other.
First lessons, observed
A specialist joins your first three periods, on call rather than in charge. Most of that time is spent on classroom logistics, not software.
Review with data
We sit down with your leadership and go through what students struggled with, chapter by chapter, and what to change next term.
It runs on what you already own
- One projector is enoughWhole-class mode needs a single laptop at the front. Everything else is optional.
- Works offlineLabs are cached and run in the browser. A dropped connection mid-period doesn't stop the lesson.
- Old hardware is fineChromebooks, shared tablets, and desktops back to about 2016. No install, no admin rights needed.
- Low bandwidth by designA lab is a few hundred kilobytes, not a video stream.
Safety, privacy and the law
- Compliant with the DPDP Act, 2023Verifiable parental consent for students under 18, no behavioural tracking of children, and no targeted advertising. Ever.
- Student data stays with the schoolStored in India, never sold, never used to train a public model. Districts can export or delete everything on request.
- No open chatbot for studentsAI hints are pre-reviewed and scoped to the current lab. Children never converse freely with a model.
- Teachers can overrule anythingEdit a lesson, hide a lab, change a score. The classroom is not run by the software.
Pricing
Priced per student per year, billed to the school. Individual teachers pay nothing, and always will — a teacher should never have to ask permission to try something.
Teacher
For one teacher who wants to try a lab next week.
- Full lab library, all grades
- Lesson plans and printable worksheets
- Up to 3 classes
- Community support
School
For a whole school, with training and support included.
- Everything in Teacher
- Unlimited classes and staff accounts
- Standards mapping and gap reporting
- Two hours of on-site teacher training
- Offline classroom mode
- Named support contact
District
For groups, trusts and government programmes.
- Everything in School
- Cross-school analytics for leadership
- Custom labs built to your syllabus
- Regional language localisation
- Data residency and procurement support
- Termly review with your team
Prices shown are placeholders. Government schools, low-fee private schools and NGO programmes are funded separately — ask us, the answer is usually yes.
What teachers tell us
“I had explained Moon phases the same way for nine years. The first time I used the simulator, a boy at the back said ‘so it was never the shadow?’ — and I realised he had believed that since Class 4.”
“What sold my staff was not the animation. It was that the worksheet came in six reading levels. That used to be my entire Sunday.”
“We were sceptical about AI anywhere near children. The thing that convinced the board was that every lesson has a teacher's name attached to it, and no student can chat with a model.”
Quotes are illustrative placeholders. Replace with cleared, attributed testimonials before going live.
Questions schools ask
Does this replace practical work in the physical lab?
No, and we would argue against any product that claimed otherwise. Children need to feel a magnet pull and watch a real beaker change colour. Simulation earns its place when it shows what the eye cannot — particles, field lines, the solar system seen from outside — or when the real experiment is too slow, too costly or too dangerous for a school lab. Our lesson plans put the real object first, every time.
How does AI actually get used, and where does it stop?
The AI drafts explanations, builds practice variants at different reading levels, translates, and flags where a class is stalling. It does not decide what a child should understand, it does not publish anything unreviewed, and it never holds an open-ended conversation with a student. Every lesson carries the name of the subject specialist who reviewed it, and reviewers delete roughly a third of every AI draft.
We have poor internet and shared devices. Is this realistic for us?
Yes. Labs are a few hundred kilobytes and cache in the browser, so once a lab has loaded it keeps running through an outage. Whole-class mode needs one laptop and a projector — in our experience one student driving the simulation at the front while thirty argue about what to change next often works better than thirty separate screens.
Is student data safe, and where is it stored?
Student data is stored in India and belongs to the school. It is never sold, never used to train a public model, and never used for advertising. We operate under the DPDP Act, 2023, including verifiable parental consent for students under 18. Districts can export or permanently delete all data on request, and we will put that in the contract.
Our teachers are not confident with technology. How much training is needed?
Two hours, then a specialist sits in on your first three lessons. The design assumption is that a teacher should be able to run a lab having opened it for the first time that morning. If a lesson needs more than five minutes of preparation, we treat that as our design failure, not the teacher's.
Can we align it to our own scheme of work?
Every lab is tagged to learning outcomes rather than to a fixed sequence, so you keep your textbook and your order of teaching. On the School and District plans we map the library to your scheme during the scoping call and show you where the gaps are.
What happens if we stop using it?
You export everything and we delete the rest. There is no lock-in clause, no penalty, and no proprietary format holding your students' work hostage. We would rather you left cleanly and came back later than stayed because leaving was painful.
See it with your own syllabus
A thirty-minute call. Bring one chapter your students find hard, and we will show you the lab for it and the lesson plan around it. If there isn't one yet, we will tell you that instead of improvising.
Give a class something to take apart.
Start with one lab in one lesson this week. Teacher accounts are free, and always will be.