For investors

Tutoring works. It has never scaled.

A student with a good private tutor learns dramatically more than one in a classroom. Almost nobody gets one, because a person watching you work costs by the hour. Klera is that person, on the iPad students already study on.

Klera on an iPad: a student's handwritten f(x) = x squared minus 6x plus 5 and its derivative, with Klera's hint asking where f prime of x is zero

The oldest result in education

One-to-one tutoring beats every other intervention we know of.

In 1984 Benjamin Bloom reported that students taught one to one performed around two standard deviations better than students taught in a class: the average tutored student outperformed nearly everyone in the classroom. Forty years of follow-up research has put the size of the effect lower, but tutoring remains one of the largest, most reliably replicated effects in education.

Bloom called it the two sigma problem: not how to make tutoring work, but how to give its benefit to everyone when a human tutor for every student is unaffordable.

What a tutor actually does

It is not the explanations. Those are in every textbook. It is watching the student work: seeing the hesitation before the first line, the erasing, the moment they go quiet, and stepping in with exactly the right amount of help. Then stepping back.

Why now

Three things changed at once.

  1. 1
    Models

    AI can finally do the maths

    Reading handwriting and reasoning through university problems went from unreliable to routine in two years.

  2. 2
    Hardware

    The pen is already there

    Students take notes and revise on iPads with Apple Pencil. The sensor a tutor needs is in their bag.

  3. 3
    Behaviour

    Students already study with AI

    The habit exists. What they use is built to answer, not to teach.

Answer machines make it worse

The fastest way to satisfy a student is the surest way to stop them learning.

The evidence

In a large field experiment (Bastani et al., 2024), students given an unrestricted chatbot did better on practice problems and worse on the exam once it was taken away. A version designed to guide rather than answer removed most of that harm.

Blind between messages

A chat window sees a student twice: when the question is typed and when the next one arrives. Everything a tutor reacts to happens in between, and is lost.

The opening

The product that wins is the one that makes students better, not the one that makes homework faster. That needs a different kind of input, and a tutor willing to hold back.

What we've built

Working today, on a real iPad.

A whiteboard that writes back

Students work by hand with Apple Pencil. Klera answers in real handwriting on the same page, and talks them through it.

A tutor that watches the pen

It reads how a student is working, not just what they wrote, and decides between a hint, a question, or silence.

A model of each student

It learns how each person gets stuck and what helps them, and gets sharper every session.

Portrait iPad showing a seven step handwritten solution of a derivative, written by Klera in gold handwriting

Why it's hard to copy

Three different kinds of difficulty, stacked.

Data

Signal nobody else collects

How a student writes, pauses, erases and asks for help only exists on a native canvas with a pressure-sensitive pen. A chat or web product cannot get it by trying harder.

Craft

Teaching, not answering

Knowing when to hold back, and writing back as natural handwriting on the student's own page, took four generations of engine to get right. It is the part a general AI company has no reason to build.

Compounding

It gets better the longer you stay

Every session sharpens the model of that student. Two months in, the help is calibrated to them. Leaving means starting from a blank page.

What decides whether a student actually learns.
 Chat AIHuman tutorKlera
Works from your handwritingNoYesYes
Sees you while you workNoYesYes
Holds back the answerRarelyYesYes
Remembers you across monthsBarelyYesYes
There at 11pm before the examYesNoYes
CostNear zeroTens of euros an hourA subscription

How the business works

Students pay first. Institutions and parents scale it.

1

Students

€9.99 a month, well below a single hour of private tutoring. Exam season is the wedge: the moment a student will pay to understand.

2

Universities

Campus licences, and insight for departments into which topics a cohort is quietly losing marks on.

3

Parents

Families already spend heavily on tutoring. Klera is a fraction of the price and available every evening.

Start where exams are still handwritten: STEM at European universities. Win one campus, then copy it.

Where we are

Written plainly.

BuiltHandwriting in and out, guided solving, exam marking, live diagrams, the learner profile
BetaFirst testers are students at Bocconi and PoliMi, on TestFlight
MilanStarting at Bocconi and PoliMi, then other EU universities with handwritten STEM exams
€0Revenue so far. Nothing is for sale during the beta
A student in a university library working on an iPad with Apple Pencil, Klera open on the screen
A tester using Klera in the library.

Ask us for the deck

Numbers, market and plan are in the deck, and we would rather walk you through the product on an iPad. Write to both founders.