A team in Paris, building a maths coach born of research work begun in 2023.
AiCole exists because a teacher cannot sit beside each of their thirty students, and because digital tools count right and wrong answers without looking for why a student got it wrong.
- Paris
- the team, with teachers and schools around it
- 2023
- start of the research work, fed by teachers’ feedback
- July 2024
- incorporation of AiCole
- 2025
- first pilot classes, with teachers
Why AiCole exists.
The problem
In a classroom, thirty students move at a single pace. The fastest wait; those who are stuck move on to the next chapter without resolving what blocked them. It is nobody’s fault: a teacher cannot sit beside each one.
What we build
Born in 2023 of research work fed by teachers’ feedback, and a company since 2024, AiCole builds support that explains from what the student knows. It reads their reasoning, names the step that blocks and helps them redo it.
Where we stand
Coach is available to students. Teacher Assistant is being built with pilot teachers. The school layer is planned for the 2027 school year.
What we believe about learning.
Four principles, and what they change in Coach. The research foundations are on the Results page.
See the journey in CoachUnderstand before answering.
A student learns nothing from an answer handed over. The aim is for them to reach the solution themselves, with the help they need, in language that speaks to them.
Read the mistake.
A mistake says almost everything about what was understood, provided you read it. A student afraid of being wrong stops trying.
Check that it holds.
What is understood today must remain usable later, mixed with other notions.
Aim for autonomy.
Students progress more when they choose and when they see their competence grow. They must be able to do without the help, notion after notion.
How we build AiCole.
Every release is designed, tested and improved with teachers and students.
One cycle, every release
Co-design
with teachers, from what is missing in class
Test
in workshops or pilots, with students
Listen
to their feedback: what helps, what gets in the way
Improve
fix, simplify, then go again
What decides, in this order
AI is a means; pedagogy decides.
- Pedagogy. The four-stage cycle (discover, assimilate, restitute, master) comes first; every feature has to find its place in it.
- Curriculum. The official mathematics curriculum, from 6e to Terminale.
- AI. Several models generate the content; exercises and exam papers pass automatic checks before they reach the student.
- Evaluation. Feedback from the pilot classes, published with its limits.
- Designed for minors. Data reduced to what is needed, no retention mechanics.
The team.
A Paris-based team. Around it, advisers, including mathematics teachers, and the teachers of the pilot classes.
Codin Segal
Artur Jurgenson
Florian R. Gayet
Calin Segal
Ludovic Winckler
Benoît de Fonds Montmaur
Laetitia de Bailliencourt
Partners and experiments.
The pilot programmeEnsemble Scolaire Notre Dame Saint Joseph
Partner school · Coach tried in class
AI for Good (ITU)
2026 use case, written by AiCole
Teacher, researcher, school, journalist: the team replies directly.
From a member of the team.