❓ Problem

The best education ever invented is also the oldest. One adult, one child, full attention. Aristotle taught Alexander this way and 2,300 years later we still have not worked out how to give it to everyone. Benjamin Bloom's famous "2 Sigma" finding was that tutored students outperform classroom students by two full standard deviations, which is the difference between a middling pupil and a top-5% one. The catch has never been the method. It has been the price. A private tutor in the US runs $25 to $80 an hour, and Kumon charges $120 to $255 a month per subject for a couple of half-hour sessions a week.

So the thing that demonstrably works is rationed by income. Roughly 21% of US students get some form of private tutoring, and fewer than 2% get high-quality, evidence-based tutoring. Everyone else gets the classroom, and the classroom is struggling. In 2024, only 31% of US fourth graders read at or above NAEP Proficient, and 40% landed below Basic, the worst showing for the bottom cohort in more than three decades. A teacher with 25-plus kids cannot fix that. Do the arithmetic on a 6-hour day and each child gets a few minutes of genuine one-to-one attention, if that.

Here is the part that should annoy you. Ages 4 to 8 is the highest-leverage window in a person's entire education. Miss phonics and number sense there and the gap compounds for the next fourteen years. And it is precisely the window where the software is worst. The category is dominated by edutainment: a fixed curriculum, a cartoon owl, a streak counter, a lot of tapping. The app does not know that your child cannot hear the "th" sound. It does not know anything about your child at all. It runs the same 400 levels at every kid on earth and calls the resulting engagement metric "learning".

✅ Solution

A voice-first AI tutor for ages 4 to 8 that teaches reading, writing and early maths at private-tutor quality, for $25 a month.

• It starts by actually assessing the child. A playful 15-minute voice session, framed as a game, that builds a live model of where the child really is: phonemic awareness, letter-sound mapping, blending, sight words, number sense, counting strategies, and the thing no worksheet captures, attention span. That model updates every single session.
• Every lesson is a story generated on the fly. Starring whatever the child is currently obsessed with: dinosaurs, diggers, princesses, sharks. Struggling with the "th" sound? Tomorrow's story is quietly stuffed with them. Nailed subtraction? The dragon now owes the knight change from a tenner. The child thinks they are hearing a story. They are being drilled.
• It is voice-first because the user cannot read yet. This is the whole unlock. A four-year-old cannot type, cannot navigate a menu, and cannot tell you what they do not understand. They can talk. The tutor listens to them read aloud, hears the stumble at the phoneme level, and intervenes in the moment, exactly as a good human tutor would.
• Parents get a weekly tutor's note in plain English. Not a dashboard. Not a chart. Three sentences: what clicked, what did not, and one thing to practise at dinner this week. This is the retention mechanic and the marketing asset in one, because it is the artefact parents screenshot and send to other parents.
• It is sold as a supplement, not a replacement. Free teacher accounts, a "share with your child's teacher" export, and positioning that makes the classroom teacher an ally rather than a threat. The category is full of companies that picked a fight with schools. Do not.

The wedge is narrow: get one five-year-old reading. The platform underneath, a persistent, per-child model of exactly how one human learns, expands into every subject, every year of primary, and eventually the most valuable dataset in education.

📊 Key Numbers

Market size

• The global private tutoring market is around $131bn in 2025, heading to roughly $143bn in 2026 at about 9% annual growth. Up-to-K-12 is the overwhelming majority of it.
• North America alone accounts for roughly $12bn of that, and the US online tutoring slice was about $4.3bn in 2024 with K-12 making up over half of demand.
• TAM: the K-12 share of global private tutoring is a $100bn-plus pool. That is the ceiling, and nobody sane underwrites against it.
• SAM: there are roughly 18 million US children aged 4 to 8 (from 73.1 million under-18s). Add the UK, Ireland, Canada and Australia and you are at roughly 25 million English-speaking kids in the target band. At the observed 21% tutoring-purchase rate, that is about 5.2 million buying households. At $300 a year, a $1.5bn serviceable market in English alone.