How it works
What happens when your child opens it.
Exploring is theirs: any question, nothing graded. Practice bends around both what they love and what they’ve shown they can do. Safety is on for everyone, at the strictest level, and when something needs your attention you’re emailed and it’s on your dashboard.
In short
Knows a lot. Tells them a little at a time.
Hazel, their AI guide, can go from sharks to black holes to how bones heal, in words their age can follow. In practice it asks first, hints next, and shows the full answer only once they’re still stuck.
Screen time that feeds their curiosity instead of eating it.
No ads, no videos, no streaks, no leaderboards, no other users. They ask, guess, explain and write, and after about twenty minutes Hazel says it’s a good place to stop.
01 · Practice
It starts with what they love. Then it learns what they’re ready for.
- They pick what they love.On their first visit they’re asked one thing (how old they are), then they tap a picture: sharks, volcanoes, space, robots. Or they ask their own question. Chosen by them, never guessed at.
- Practice is shaped around it.Math, science, reading, English, writing and social studies, written after they’ve shown what they’re into and aimed at their grade’s learning goals. Math numbers come from our code, and every retelling is solved separately to prove the answer didn’t change. Questions in other subjects are written by the AI and used only if a second, independent AI reaches the same answer, the interest fits naturally and the safety check passes; otherwise they get a question from our hand-checked set.
- They think it through first.The guide opens every problem with a question, and gives a strategy that suits it, not just “try again”. The fresh problem they do on their own appears once they’ve had a go. On multiple choice they say why they picked it, so a lucky guess isn’t counted as learning.
- It starts easy, and climbs only when they’re ready.Every new skill starts at a level chosen for their grade. Right first time with no help, twice? It comes back a step up. Wrong, or needed help (from a button or in their own words)? It stays or steps back down. Mastered? It rests for a month, then comes back to check it stuck. A harder problem is always earned, never sprung on them.
- Stuck is never a dead end.If it gets too hard, “Try an easier one” swaps in a gentler version, with no penalty. The guide watches for frustration too: it backs off to a smaller step of the same problem, and if they’re still stuck, it shows them properly and gives them a fresh one.
- It won’t just hand over answers.“Just tell me” with no try behind it gets the next small step, not the answer. Every problem ends with a fresh one to check it stuck, and a few days later it comes back to see if it lasted.
- It has an end.A session is three problems and some exploring. After that, or about twenty minutes, it says it’s a good place to stop, and shows them what they did and the question they saved for next time.
RatiosMinecraftLevel up
In a Minecraft farm, the ratio of sheep to cows is 3 to 5. There are 24 sheep and cows altogether. How many sheep are there?
just tell me the answer
I’ll get you there. One group has 3 sheep and 5 cows. How many animals is that in one group?
Five subjects, pitched at their age
- MathFrom adding, shapes and measuring to fractions, ratios and simple equations.
- ScienceLiving things, Earth and space, forces and matter.
- ReadingReal paragraphs: main ideas, clues, and reading between the lines.
- EnglishGrammar, punctuation, what words really mean, and their own writing, read by the guide rather than marked.
- Social studiesMaps, communities, and how people lived long ago.
For roughly Grades 2 to 8, aimed at grade-level learning goals. It isn’t a complete curriculum, and exploration covers anything they’re curious about.
02 · Safety
A place to wander that can’t lead anywhere dangerous.
Children use this, so safety isn’t a setting you have to find. It’s on for everyone, at the strictest level available, and when in doubt it stops.
What they can’t get to
However they ask.
- Weapons, explosives and poisons
- Drugs, and how to get high
- Ways to hurt themselves, anyone else, or an animal
- Anything sexual
- Gore and graphic violence
- Hate, and cruel jokes about people
- Hacking into other people’s accounts
- Strangers, secret friendships, and sharing where they live
No internet, at all. It can’t search the web, open websites, or show pictures or videos, and any web address in a reply is removed in code before they see it. It only knows what it learned; it can’t go and look.
How it’s enforced
- Clever wording doesn’t get around it.“Ignore your rules.” “My dad said it’s fine.” “Pretend you’re my grandma.” “It’s for a story.” Misspellings, fake system messages, and slow step-by-step approaches. We test it the way a determined kid would, against the live system, at ages 8 and 11.
- Every message is checked, both ways.What your child sends and what the AI says back are reviewed before anyone sees them. Replies are never shown half-written.
- Four layers, not one.An instant word check that works even if the AI is down. Google’s child-safety protections, which can’t be switched off. Every adjustable filter at its strictest on everything written for them. And a reviewer that can tell a child testing a boundary from a child who needs help.
- If the check can’t run, nothing is shown.A broken safety check stops the conversation. It never waves a reply through.
- A no that fits, and a way back.Typing an address gets “let’s keep that private”; asking to keep secrets gets “that’s one for a grown-up”; an off-limits topic gets a friendly no. Each one offers a way back to what they were exploring.
- Real questions still get real answers.War, death, volcanoes, why people take drugs: answered honestly and briefly, gentler for younger children, without graphic detail, and pointed back to you. A child who gets stonewalled goes and asks somewhere worse.
- If they ask about hurting themselves, you’ll know.Questions like that, or anything suggesting they may not be safe, get a kind reply pointing them to a trusted adult, an alert on your dashboard straight away, and an email telling you to look (never what was said). A second pair of eyes, not emergency monitoring.
- Nobody else is in here.No chat with other users, no public profiles, no streaks, no leaderboards, no ads.
Yours to control
- Every account is approved by hand.Signing up only files a request. Nothing works until it’s approved.
- You’ll know when something was turned away.The session report tells you how many messages the safety check declined, and what kind (personal details, secrets or strangers, or an off-limits topic), never what was typed.
- Only grown-ups sign in.Your child uses it under your account: no child email, no child login.
- A grown-up PIN keeps the parent pages away from children.Your dashboard, reports and settings ask for it; the child’s own screens never do.
- One account for the whole family.Add up to 6 children, each with their own sessions, practice and section of the Sunday report, and your agreement recorded for each one.
- You choose how long we keep it.30 or 90 days, then records delete themselves.
- Export or delete, any time.Every child’s learning records, from the settings page: download them all, or delete one child’s or everyone’s.
- You can set where practice starts.Choose a skill for each child in settings, or let it choose from their record.
- Talking instead of typing is your call.Off unless you turn it on, child by child. Their device turns speech into text; we only ever receive the text.
- Daily limits are built in.Up to 10 sessions and 40 messages a day for each child (20 and 120 for the whole family), so it can’t swallow the evening. Each family also has a monthly AI allowance, about CA$25 of AI, which we pay for. If it’s ever used up, sessions pause until next month.
- Straight answers about data.Stored in Montreal, with the AI called in the same region, and a plain account of what that does and doesn’t cover in our privacy notice.
03 · The rules we hold ourselves to
Built to be trusted with a child.
- It helps them understand, not copy.Hints before answers, and a fresh problem after.
- No streaks, no leaderboards, no nagging.Nothing designed to make a child anxious about stopping.
- We never label your child.No “gifted”, no “behind”, no comparison to anyone else.
- Hard questions get honest answers.Age-appropriate, brief, and pointed back to you.
- Honest when it isn’t sure.It’s an AI and can get a fact wrong. It’s told to say when scientists aren’t sure, and we tell kids to check big facts with you.
04 · Questions parents ask
Before you ask for access.
- What does it cost?Free in the pilot, until at least 31 January 2027, with at least 30 days’ email notice before that changes. Then CA$29 a month or CA$249 a year per family, or CA$19 a month while you stay subscribed if approved in the pilot. You pay only if you choose a plan; without one, sessions pause. Later families get 14 days free, no card. Each family also has a monthly AI allowance, about CA$25 of AI, which we pay for. If it’s ever used up, sessions pause until next month. Cancel any time in Settings.
- How long does approval take?Chris reads every request himself and replies by email within two days. Until then the account can’t be used.
- Who is Chris?A dad who built this for his own son, and the only person behind it. He approves every family, reads every message sent through the contact page, and replies to them himself.
- What does my child see?On their first visit, one question: how old are you. Then picture cards (sharks, volcanoes, space) and a box for their own question. They explore with the guide, then practise three problems written around what they picked. No ads, no other users, no internet.
- Why not just use ChatGPT?An open chatbot hands over answers and has no idea it’s talking to a child. Here, “just tell me” gets the next small step, not the answer; every message is checked both ways before anyone sees it; there is no web to wander off into; and on Sunday you get the story.
- Can it be wrong?Yes. It’s an AI. Math answers are set by our code, not the AI, but in exploration it can get a fact wrong. It’s told to say when scientists aren’t sure, and we tell kids to check big facts with you.
- What data do you keep, and where?A nickname (not a full name), age and grade, their sessions and the practice they did. Kept 30 or 90 days, your choice, then deleted. Stored in Google Cloud in Montreal, with the AI called in the same region. No ads, nothing sold, and your child’s words are never used to train AI. Download or delete it all from settings.
The details are in the privacy notice and the terms, and the prices on the pricing page. Anything else, ask us.
See it with your own child.
A private pilot for a small number of families. Every account is approved by hand.