Robotics they can hold. Physical AIthey can teach.
The world is moving from digital AI to Physical AI and intelligent robotics — part of a global tech
shift projected by the World Economic Forum to create
170 million new jobs by 2030.
Is your child ready?
Four Products,
One Path
Design it
MOONTINKER
Give it a body
MOONVERSE
Control it online
MOONIOT
Deploy on real robot
PHYSICAL AI
What They
Do Together
each step takes more than one product
Embodied AI Design
Give the design a body
Synthetic Data Generation
Make the training data
Sim-to-Real Validation
Test it in the real world
Real-Hardware Deployment
Bring it to real hardware
Skills They Build
Logical thinking
Algorithmic thinking
Debugging and patience
Circuits and physics
Creative problem solving
Measuring and testing
Where They Show What They Built
Show & Tell
Every quarter · they present what they built
A
M
J
MoonBattle
Once a year · every child, judged by industry
S
R
K
THE ROADMAP
From robotics and coding to AI and machine learning
Six stops, in order, each one only possible because of the one before it. You cannot train a model on
data you never collected, and you cannot collect data from a robot you never got working. MoonTinker is
the workbench at every stop.
Hover or tap a stop to see which products are in play.
Each stop is tagged with the tier it belongs to: T0 is ages 7–8, rising in two-year bands to T3, with T4
and T5 open from 13.
ROBOTICS & CODINGAI & MACHINE LEARNING
WHAT EACH PRODUCT IS FOR
MoonTinker — design the circuit, write the program, debug it, flash the
board
MoonVerse — give it a body and a world, test the behavior
MoonIoT — online, controlled
Physical AI
WHAT THE PLATFORM DOES WITH THEM — EACH ONE TAKES MORE THAN ONE PRODUCT
Embodied AI Design
Synthetic Data Generation
Real-Hardware Deployment
Sim-to-Real Validation
MoonTinker
MoonVerse
MoonIoT
Physical AI
Six Stops on the MoonTinker Road
MoonTinker spans the whole road. Every later stage still begins by designing a circuit and writing the program in it — AI is added on top of that work, never instead of it.
1T0 · 7–8
Circuits and blocks
Wire a working circuit and make it do something on command.
Real components, real wiring rules. The program is built by snapping blocks together, so there is no syntax to fight yet.
MOONTINKER
2T1 · 9–10
Python and algorithms
Read and write a program as text, and explain the steps in it.
The same project switches from blocks to Python. Loops, conditions, functions, and the first real algorithms — line following, thresholds, timing.
MOONTINKER
3T1 · 9–10
Behavior and testing
Judge a program by what it does, not by whether it runs.
The circuit is given a body and a world. Now the question is whether the garden actually gets watered and whether the car actually holds the line.
MOONTINKER + MOONVERSE
4T1 · 9–10
Real hardware
Move their own program off the screen and onto a board that moves.
A USB cable and one button in the toolbar. The gap between what the simulator predicted and what the robot does is itself a lesson.
MOONTINKER
5T2 · 11–12
Connected devices and data
Put a device online, control it from a phone, and read what it reports.
Sensors start producing a stream of numbers over time. Recording, plotting and questioning that stream is the step that makes machine learning possible.
MOONIOT
6T4 · 13+
AI and machine learning
Train a model on data they produced, then measure honestly how well it learned.
They drive the robot, the app records it, and a small neural network learns to copy them. Then they break it on purpose — noise, dim light, low contrast — and watch where it fails.
PHYSICAL AI
So where does MoonTinker fit?
It is not one stop on the road — it is the bench the whole road is built on. Stops 1, 2 and 4 happen entirely inside MoonTinker: the circuit is designed there, the program is written there in blocks or Python, it is debugged line by line
there, and it is loaded onto the real board from its toolbar. MoonVerse, MoonIoT and Physical AI each add something on top of a MoonTinker project — a world to test in, a network connection, a trained model — but the circuit and the
code underneath stay exactly where the student built them.
It starts with a circuit a child wires themselves and a program they can read, and ends with a robot that learned to drive by watching them do it. Four products carry them across that gap one step at a time — and
the mistakes they make on the way do most of the teaching.
BEYOND THE ROBOT
What it builds that lasts
The robot goes in a cupboard eventually. These are the parts that do not.
Logical thinking
If this, then that — expressed precisely enough that a machine will do it, which is a far stricter test than saying it out loud.
Algorithmic thinking
Breaking a goal into ordered steps, then noticing which steps repeat and turning them into a loop.
Debugging and patience
Their program will not work the first time. Finding their own mistake, calmly, is the habit this builds.
Circuits and physics
Voltage, current, resistance, sensors and motors — learned by wiring them, and by seeing what happens when the wiring is wrong.
Creative problem solving
There is no single right answer to "make the plant get watered." There are many, and they get to argue for theirs.
Measuring and testing
Change one thing, run it again, compare — the beginning of real experimental method.
FOR PARENTS
This is the actual screen your child works on
Not a cartoon version of engineering. The same canvas, the same debugger and the same live classroom that mentors use every session.
They wire it themselves
Components come off a searchable shelf and get wired up by hand. The simulation runs as they build: here the ultrasonic sensor is reading 41.26 inches and the micro:bit beside it is showing
what the program put on its display.
Nothing is pre-assembled, so a wire in the wrong hole behaves exactly as it would on the desk — which is the point.
The circuit canvas, with the component palette open.
Paused at a breakpoint, every variable visible.
And they find their own fault
This is a real debugger, not a hint button. The program is paused at a breakpoint on line 19, and every variable in it is listed with its current value while the circuit sits there frozen mid-run.
A child who can stop a program and ask "what is this variable actually holding right now?" has learned something that outlasts any particular language.
Their mentor is watching the same screen
In a live session the mentor sees every student's canvas at once and can drop into any one of them to guide, correct or demonstrate — so a stuck child is noticed in seconds rather than at the end of the hour.
The same view is how work gets reviewed afterwards: your child's exact circuit and code, opened and graded by the person who taught them.
The mentor's live view of a class in progress.
THE SHELF
117+ real components, not blocks that pretend to be them — and growing
Every part behaves the way the physical one does — the LED has a forward voltage, the resistor limits current, the servo has a travel range. This is a sample of the shelf, and the shelf keeps getting longer.
LED
Resistor
RGB LED
Push button
Potentiometer
Light sensor
Buzzer
Ultrasonic
IR array
Servo
DC motor
Battery
Breadboard
NodeMCU
…and a hundred more: displays, keypads, temperature and humidity sensors, accelerometers, relays, micro:bit and MoonBit boards. New parts are added as the programs grow, and they arrive in the editor your child already has open — nothing
to download.
MOONVERSE
Then the circuit gets a body and a world
More than ten of them, and growing. The question stops being "does the program run?" and becomes "does the garden actually get watered?" — which is a much harder, much more interesting question.
The garden
A moisture sensor, a sprinkler, and a plant that will genuinely dry out if the logic is wrong.
The line follower
Three infrared sensors, a steering servo, a drive motor, and a line the car keeps losing.
Obstacle detection
An ultrasonic sensor on the front, walls, barrels and rocks in the way — stop in time, or steer around.
The home
Keypads, lights, fans and appliances — a house that arms, disarms and responds.
The football pitch
Robot players, a keeper and a goal. Strategy expressed entirely in code.
The smart garage
A car pulls up, a sensor spots it, and a servo lifts the door — then closes it behind them. Simple to describe, surprisingly fiddly to get right.
Plus a control panel, smart builds, a moon garden, a drone and a self-driving car — with new worlds added as the programs grow.
EMBODIED AI DESIGN
The AI gets a body — and your child designed it
In most AI lessons the model is a thing in a browser tab with no hands. Here it is wired to a machine the student built: they chose what it can sense, where those sensors point, and what it is able to move. The model is fitted to
that machine, and if they change the machine, it has to learn again. That is what makes it engineering rather than a demo.
WHAT IT SEES
576
numbers, every frame
a 32 × 18 strip of the road
27,696
HIDDEN
48
it finds the edge
784
HIDDEN
16
it finds the turn
34
WHAT IT DOES
2
steering · drive
28,514 values in all — the whole of what the model knows. Small enough that a student can be told, truthfully, what every part of it does, and small enough to train in the browser in seconds.
They choose the senses
Three infrared sensors looking down, or an ultrasonic looking ahead? Mounted where? A machine can only act on what it can perceive, so this is the first real design decision — and the first one that can be wrong.
The model learns the link
It is not told the rule. It watches 1,400 frames of the student's own driving, twenty a second, and works out for itself what a road looks like when you should be turning left.
They choose the body
A steering servo and a drive motor — so the model has exactly two numbers to say. Put bigger wheels on it, or move a sensor an inch, and the old model is wrong. It has to learn the new body.
THE LEARNING PATH
Six tiers, ages 7 to 18 — with several programs in every tier
A child joins at the tier that matches their age and moves up roughly every two years. From thirteen the top two tiers open up, and a student enters them with what they have already built rather than on their birthday. Each tier holds
more than one program, so there is room to go sideways and deeper before going up — and nothing is dropped on the way: a student training a model is still designing the circuit and writing the logic they learned at seven. The hardware
kit is not an upgrade or an add-on: it comes with every program, in every tier.
Click a tier to see what it holds
The hardware kit — in every tier, with every program
MoonTinkerMoonVerseMoonIoT
Physical
AI
Each tier is a taller stack, not a different subject. MoonTinker is at the bottom of every one of them, and the hardware kit runs underneath the lot — every tier, every program, in it.
T0
7–8
WHERE EVERYONE STARTS
MoonTinkerMoonVerseHardware kit
First circuits, first programs
Wiring real components and making them do something on command, with programs built by snapping blocks together — no typing, no syntax. The circuit is then given a body in a simulated world, so they can see what their logic
actually does.
SEVERAL PROGRAMS IN THIS TIER
The hardware kit comes with every program in this tier
T1
9–10
BLOCKS BECOME PYTHON
MoonTinkerMoonVerseHardware kit
Python, algorithms and real hardware
The same project switches from blocks to Python and keeps going. Loops, conditions, functions and the first real algorithms — line following, thresholds, timing — tested as behavior rather than code, then flashed onto a physical
board over USB.
SEVERAL PROGRAMS IN THIS TIER
The hardware kit comes with every program in this tier
T2
11–12
DEVICES GO ONLINE
MoonTinkerMoonVerseMoonIoTHardware kit
Connected devices and data
Circuit design with the physics behind it, and devices that connect, sense what they sense and are controlled from a phone. Readings are stored, streamed and plotted over time — which is exactly what the next tier needs.
SEVERAL PROGRAMS IN THIS TIER
The hardware kit comes with every program in this tier
T3
13–14
DEPTH, NOT BREADTH
MoonTinkerMoonVerseMoonIoT
Harder algorithms, real physics
Advanced logic and algorithms, circuit and control with the physics actually accounted for, and connected systems that log what they sense over time. This is the tier that builds the groundwork AI needs — and the data it will be
trained on.
SEVERAL PROGRAMS IN THIS TIER
The hardware kit comes with every program in this tier
T4
13+
AI STARTS HERE
MoonTinkerMoonVerseMoonIoTPhysical AIHardware kit
Training their first model
They drive the robot. The robot records what they did, and a small neural network learns to copy them. Then they measure how well it learned and break it on purpose — noise, dim light, low contrast — to find out exactly where it fails.
SEVERAL PROGRAMS IN THIS TIER
The hardware kit comes with every program in this tier
T5
13+
END TO END
MoonTinkerMoonVerseMoonIoTPhysical AI
A project they own completely
Designed, built, deployed on hardware and defended: the whole arc in one piece of work, at a standard they can put in front of anyone.
SEVERAL PROGRAMS IN THIS TIER
The hardware kit comes with every program in this tier
WHAT ARRIVES AT HOME
Three robotics kits per tier, posted to your door
The shelf on screen is only half of it. Real boxes of parts arrive through the year — one in the first month, one in the fourth, one in the eighth — so the hardware turns up at the point in the tier where a child is ready to build
with it, rather than all at once on day one.
T0
7–8
MOONSPARK
1ST MONTH
Moonspark Builder
4TH MONTH
Moonspark Explorer
8TH MONTH
Moongarden
Moonspark Innovator
T1
9–10
MOONBIT
1ST MONTH
Moonbit Builder
4TH MONTH
Moonbit Explorer
8TH MONTH
Moonrover
Moonbit Innovator
T2
11–12
MOONPI
1ST MONTH
Moonpi Builder
4TH MONTH
Moonpi Explorer
8TH MONTH
Moonkart
Moonpi Innovator
Tiers 0 to 2 shown. Every tier ships its own kits on the same cadence.
WHAT STOP SIX ACTUALLY PRODUCES
A result they measured, not a demo they watched
A student drives the car, a small network learns to copy them, and then they deliberately spoil the camera — adding noise, dimming the light, flattening the contrast — and measure what the model does. Noise and brightness barely move
it. Contrast breaks it. Finding out which thing breaks your model, and why, is the lesson.
The MoonKart pilot degrades gracefully — except on contrast
Mean steering error as the camera is degraded, measured on 300 held-out frames the pilot never trained on. 1.0 would be full lock, so 0.012 is a pilot holding its line. Simulation only — no hardware involved. Hover a point to read
its value.
Sensor noise
grey levels, σ
Brightness shift
grey levels
Contrast
× of true contrast — the fragile axis
Measured on 300 frames the model had never seen. Produced by a student-scale network of 28,514 values — small enough that they can be told, truthfully, what every part of it does.
FOR STUDENTS
Here is what you are looking at
Five things on your screen. You will use all of them in your first project.
117+ parts on the shelf
LEDs, resistors, sensors, motors, displays, boards — search the palette and drag what you need.
Blocks ⇄ Python
Build it in blocks, flip to Python, and it is the same program. Nothing is thrown away when you move up.
A real debugger
Breakpoints, step, and a live list of every variable. Use it the moment something behaves oddly.
10+ worlds to test in
A garden, a line-following track, a home security panel, a football pitch, a drone, a self-driving car, and more.
One click to hardware
Connect the board over USB and flash it straight from the toolbar. No extra software to install.
Your mentor, live
They can see your canvas and join it to help. Ask early — being stuck is not a thing to hide.
Three things that make wiring quick
01
Blue guides line things up
They appear on their own when a wire is level with a node or another wire.
02
HoldShiftto lock the angle
Lines snap to clean horizontal, vertical and 45° runs instead of wobbling.
03
Tabcycles, Enterconfirms
The gray dotted line is MoonTinker guessing your route. Tab to see other guesses, Enter to take one.
Drawing a wire on the real canvas — watch the cursor route it.
How an assignment works
Your teacher sets a project. It appears on your grades page and moves through three states, so you always know where it stands.
DueSubmittedGraded
Once it says Graded, open it to read your score and what your teacher wrote. Your work lives under My Projects, and your scores under My Grades.
What you need to start
A laptop or desktop running Windows or macOS. Not a tablet or a phone.
The Chrome browser. MoonTinker is built and tested against it.
Nothing to install — MoonTinker runs in Chrome, and loads programs onto your board from there too.
Your hardware kit, which comes with the program.
ABOUT ONCE A QUARTER
Show & Tell
Every few months each student presents what they have learned. They take an audience through the project they have been working on — what they set out to make, how it works, and what they worked out along the way.
Building something and being able to explain it are two different skills, and only one of them gets practiced by writing more code.
Ethan Savelsberg
built the Sensor stick
“I made this circuit to help people who cannot see very well.”
PROTOTYPE VERIFIED
Build
Code
Test
Learn
Their own project, demonstrated
Not a test and not a recital. The student shows the thing they built actually running, walks through the decisions behind it, and answers questions about it.
Parents and peers are the audience
They present to their own group and to parents. It is a real audience and a real deadline, which is most of what makes it work.
Feedback, not a winner
A mentor gives structured feedback against a rubric covering technical skill, creativity and presentation, and that goes to parents as a progress report. Nobody is knocked out.
What they actually bring
PROTOTYPE VERIFIED
Sensor stick
“
I made this circuit to help people who cannot see very well. It is designed to be a cheaper alternative to a walking stick, costing around $8 to $10.
”
“
My project is a password-protected lock that shows how security systems work. If the right code is entered, the lock opens; if not, it stays locked.
”
“
When the logo on the micro:bit is pressed, it shows a traffic light countdown... Once the green light appears, the race begins. This helps everyone start at the same time and makes the race fair.
”
“
When you shake the Micro:bit, it chooses a random number from 1 to 10. I used the random module to make the numbers random and the music module to play sounds.
”
Show & Tell runs as part of the Innovator Program, not as a separate sign-up. Each student keeps a portfolio page, and these six are a handful of the projects on it.
Once a year every child takes their own project in front of judges from outside Moonpreneur — engineers and founders from the technology industry, and people who invest in it — and makes the case for it.
THE ROAD TO THE FINALE
1
Qualifying rounds
Through the year, grouped by age
2
The finale
Judged by engineers, founders and investors
and every other child
Every child gets a turn
It is not a selection. Each student takes part once a year, grouped by age so that a nine-year-old is not up against a fifteen-year-old. Ages run from 7 to 16.
Rounds, then a finale
Qualifying rounds run through the year and the finalists meet at the finale. Recent editions have run online, so taking part has not meant traveling.
Judged on more than the build
The criteria cover the problem it solves, whether the prototype actually works, and how well it is explained. There are prizes, but the panel is the point.
The format has changed from year to year — it has run as a pitch competition, as a conference with competitions inside it, and most recently as a digital event. Check the site for the current edition, who can enter and what it costs.
Students sign in through their Moonpreneur account, and their saved projects, submissions and grades are tied to that account rather than to one computer. Their teacher and mentors can see the work they submit.
How Moonpreneur handles personal information is set out in full in our published policies:
The full architecture, for anyone who wants to see it. Solid lines are built and working; dashed lines are planned.
MOONPRENEUR
The Moon Physical AI Ecosystem
From digital conception to physical realization
drives the scene, sensors write backsends the program to a real boardflashes a trained pilotreturns the measured gapreads and writes
Solid = shipped and demonstrable today · Dashed = planned · October 2026
1
A real AI, trained on your child’s own driving.Not a demo model shipped with the product — one they create, in the browser, in seconds.
2
The loop closes.The same pilot runs in simulation and on the robot, and the difference is measured every run rather than claimed.
3
One click from screen to desk.The web app loads the program onto a real board; the phone app then controls it and reads its data.
4
One project, one store.Circuits, scenes, recordings, trained pilots and results all belong to the same child, in the same place.
5
Assessment across the whole thing.Wherever the work happens, it can be submitted, reviewed by a teacher and reported to you.
Frequently Asked Questions
Answers to common questions about MoonTinker, your child’s learning, and what you need at home.
Does my child need to know how to code already?
No. The learning path starts with circuits and visual blocks.
Children move into Python and algorithms as they progress,
building on the same projects rather than starting over.
Is this just a game, or real engineering?
The projects use real components, circuits, code, sensors and motors.
Children test their designs in simulation, debug their own programs,
and load them onto physical hardware.
What do we need at home?
A Windows or macOS laptop or desktop, the Chrome browser,
and the hardware kit that comes with the program.
MoonTinker runs in the browser; a tablet or phone is not
a substitute for the computer.
When does AI come into it?
AI and machine learning are the sixth stop on the roadmap.
Children first build a foundation in circuits, programming,
behavior, real hardware and connected devices, then train a model
and compare its performance in simulation and on the robot.
How do I know how my child is doing?
Assignments move from Due to Submitted to Graded.
Students can read their scores and teacher feedback in My Grades.
Show & Tell also gives them a chance to demonstrate their own
project to parents and peers, with mentor feedback shared as a progress report.
Can work be lost?
Saved projects, submissions and grades are tied to your child’s
Moonpreneur account rather than to one computer.
They can find saved work in My Projects when they sign in.
If something appears to be missing, their teacher or mentor
can help them check it.
See a class before you decide
The quickest way to understand MoonTinker is to watch a child use it for forty minutes.