A vision document for parents · every claim is sourced
Raise a mind that can direct the machine — not one that leans on it.
KiDI is a cognitive companion for children ages 3–8, built from a century of learning science and one founding rule: AI works for your child, and never in place of your child's thinking.
01 Why this, why now
Search and first drafts just became a commodity.
Education has absorbed the printing press, the calculator and the internet. Each time, the debate was the same: will this technology weaken young minds, or free them to think at a higher level? AI raises the same question, at a speed and scale we haven't seen before.
Generative and adaptive AI can now do the things that once defined "being educated": recalling facts, drafting essays, solving equations, explaining almost anything on demand. That doesn't make education obsolete — it makes the old definition of education obsolete. When lookup and first drafts are automated, human value shifts to what AI still can't do well: framing the right problem, exercising judgment, creating something genuinely new, and working with other people.
A handful of countries are already closing that gap — Singapore's EdTech Masterplan 2030, South Korea's national AI curriculum, Estonia's teacher tooling, and the UAE's K–12 AI programme all treat AI literacy as core, not optional.6 Most systems are still debating whether to allow AI in the classroom at all. That gap — between what children will need and what most schools currently teach — is the one KiDI is built to close, starting in early childhood rather than waiting for middle school.
02 The science we didn't invent
We just encoded a century of learning research into software.
Before designing a single activity, we studied what a century of research and classroom practice already tells us about how young children learn best. Three findings shaped every screen in KiDI.
Montessori-style choice beats fixed-pace instruction
A 2006 Science study on lottery-admitted Montessori pupils found stronger academic and social outcomes by age five.2 A 2025 national RCT of 588 children across 24 public Montessori schools confirmed it — better reading, memory and executive function at kindergarten, for thousands less per child.3
Retrieval practice builds durable learning
Actively producing an answer — not just being shown one — builds far more durable memory than passive review, even when study time is identical.4 That's why KiDI runs on short, repeated, low-stakes questions instead of long passive videos.
Parental involvement helps — the right kind
A 2023 meta-analysis found supportive involvement (helping, encouraging, discussing) predicts higher achievement, while intrusive, over-controlling involvement predicts lower.5 That's the difference between a weekly report and a scoreboard.
What we're honest about: we use encouraging, effort-focused language ("growth mindset" framing) in KiDI's feedback because the evidence leans positive — a 2019 national experiment found modest, real gains, concentrated among lower-performing students10 — not because the effect is settled science. Earlier meta-analyses found a weaker average effect,10 and we'll keep tracking the research rather than overstate it.
03 Our founding principle
Use AI. Learn AI. Never hand it the thinking.
A child who only consumes AI-generated answers is trained to be a consumer of intelligence. A child who's taught how AI works, where it's reliable and where it fails, is trained to direct it.
We borrow Singapore's own four-part framework for thinking about AI in a child's education, because it's the clearest formulation we've found in any national policy document.6
What it is, how it works
Age-appropriate explanations of the systems behind everyday tools — building the intuition that technology is a system to understand, not magic to consume.
As a tool, with judgment
Guided, bounded interactions — never an open blank page that could overwhelm a 3–5 year old.
As a personalised tutor
The app adapts difficulty, pace and content to the child — without ever generating the answer for them to copy.
Think without it, too
The child always performs the cognitive act — choosing, recalling, comparing, explaining — so the capacity to think stays theirs.
The real risk isn't too little AI
It's AI used as a substitute for effort. Singapore's own ministry has warned that premature, unstructured AI use in early years risks "cognitive atrophy" — foundational skills weakening before they're built.7
No shortcuts to the answer
KiDI never answers a question for the child. The app offers choices, listens to responses, and adapts — the retrieval effort always belongs to the child.
Guardrails are a feature
Every activity is capped, paced, and paired with a parent view — designed, in the same language Singapore uses, "to minimise cognitive delegation and AI dependency."6
04 Inside KiDI
Five cognitive domains, not five school subjects.
The app opens on one deliberately simple question — "What shall we practice today?" — and a short menu. The range of options is ours; which one comes next is the child's, echoing the Montessori principle of structured choice.2
Tap & Find
Perception & vocabulary — matching a word, a sound and an image, the bedrock of language and categorisation.
Memory
Working memory & retrieval strength, via the paired-matching format long used in developmental psychology.
Attention
Selective attention & error detection — precursors to the visual discrimination that later supports reading and math.
Logic
Pattern completion & sequencing — an early, age-appropriate precursor to the systems thinking employers rank as rising fastest.1
Explore & Discovery
Curiosity & general knowledge — the "explain it back" prompt is itself a form of retrieval practice.4
A visual progression built for every age
A 3-year-old and an 8-year-old don't perceive an image the same way. Rather than one illustration style for every age, every KiDI subject exists across four visual stages, scaffolding recognition from simple shapes to full real-world detail — and always keeping the same pose and character, so nothing feels like a break.
05 Parents at the center
A five-year-old can't tell you if they're really learning. So the app does.
Every week, KiDI writes a plain-language summary a parent can read in under a minute — built to start a precise, pressure-free conversation at home, not deliver a generic "well done."
Encourages better habits, not just praise
If 34 tasks all happened in one Sunday session, the report says so — and explains that short, spread-out sessions help new skills "stick" better than one long one.
Surfaces gaps, not only wins
Skills marked "not tried yet" sit right next to a perfect streak, so the parent's next step is obvious, not vague.
Gives precise, actionable numbers
Strongest category and weakest category, with real success rates and response times — never a single opaque score.
Suggests a real-world extension
One low-effort question to ask at dinner — turning an on-screen discovery into an actual conversation.
06 For teachers
For schools · piloting with partner classroomsThe same rule as the child's screen: KiDI proposes, a teacher decides.
A teacher already carries five classes' worth of judgment about who needs more practice, and where. The teacher layer isn't a new curriculum to deliver — it's a companion account that turns what the app already sees at home into something usable at school: one roster, one view of the official programme, one proposed plan for the year.
Everything below follows the same founding rule as the child-facing app itself: KiDI narrows and suggests, a human decides.6 No student is added, leveled, or scheduled onto a roadmap without a teacher confirming it first.
Link a class in under a minute
A teacher creates a class the way they'd create a folder — name, grade, and the official programme it follows (more on that in step 3). KiDI generates one join code and QR, valid for that class only. Each student enters it once; from then on the app already knows their class, their teacher, and the programme they belong to — no student email required.
- One join code per class, no student email or account setup needed
- Students join from their own device, or a shared classroom tablet, in a few taps
- A live counter shows who's joined before the first activity even starts
Every student, one roster — organized by class
Teachers running more than one class switch between them with a single tap. Each row mirrors the same plain-language summary a parent already sees at home: age, the domains they've tried, and how recently they were last active — so a teacher can spot who's ahead and who's gone quiet in seconds, not in a spreadsheet.
- Multiple classes under one account — no separate login per group
- Same five cognitive domains a parent sees, not a separate teacher-only score
- Nothing is hidden from the matching parent account — visibility runs both ways
The official programme, mapped next to KiDI's content
At setup, a teacher selects the official early-years or primary programme their state or ministry has published for that grade. KiDI tags every activity in its five domains against the matching strand of that programme — the same kind of alignment Singapore's own Student Learning Space already runs for math and geography.6 The result is a coverage view, not a guess: which strands the class has already touched this term, and which haven't come up yet.
- Programme selected once, at class setup — the official text, not a generic label
- Every KiDI activity tagged to its matching official strand
- A coverage gap is visible before a term ends, not after
A year roadmap KiDI drafts — the teacher signs off
Once a class is linked, its roster is in, and a programme is selected, KiDI proposes a term-by-term roadmap for the year. The suggestion accounts for the age spread in the room, each student's starting point per domain, and the terms left before the programme's own checkpoints — pitched at the group's zone of proximal development rather than one fixed pace.8 It is only ever a draft: a teacher can reorder, swap, or delete any block, and nothing becomes the class's live plan until they hit Validate.
- Every suggested block states why it's placed there — which strand, which gap it closes
- Fully editable before validation — reorder, swap, or remove any block
- Nothing reaches a student's device as an assignment until a teacher validates it
KiDI can read a room full of different levels. It still takes a teacher to decide what that room needs next.
07 Screen time, respected
We're not trying to get your child to spend more time on KiDI.
We want them to learn more in the screen time you've already chosen to give them. Health authorities differ by country, but converge on one principle: reason in daily minutes, and screens should never replace sleep, movement or human interaction.9
| Age | Most common guidance |
|---|---|
| 0–2 yrs | No screen time, aside from occasional video calls |
| 2–5 yrs | ≤ 1 hr/day in most countries; 30–40 min/day in France |
| 5+ yrs | Often ≤ 2 hrs/day of leisure use; focus shifting to content quality over a single number |
A daily budget, never a weekly average. 60 minutes today plus 0 tomorrow is not the same as 30 minutes every day.
How KiDI signals time already spent today:
The final decision always stays with the parent — the app never silently locks a child out. Sessions run 5–10 minutes, with no forced streaks, no autoplay, no infinite scroll, and no notifications aimed at the child.
08 What's next
Closing the loop between school and home.
Everything above describes KiDI as it exists today: a cognitive companion for home, with a piloting classroom layer for teachers, built to sit alongside school rather than replace it. A later version goes further still, with the backing of learning-science labs and pilot schools.
The idea is simple: the app would revisit that day's or week's classroom topic and test it again — not by reproducing the workbook, but by reshaping it into the short, playful, retrieval-based format KiDI already uses. A lesson on farm animals, a number concept, or a history topic becomes a five-minute activity at home that same evening.
Built with teachers, never instead of them.
Singapore's own national Student Learning Space already connects to the curriculum for math and geography.6 That's the level of integration we're aiming for — adapted to early childhood.
Built for parents who want proof, not promises.
Join the early access list to hear when KiDI opens — and get the full sourced vision document.