Winny AI builds educational AI ready to support learners in Comoros

Five selected education recommendations

Five selected recommendations from the ministry’s 2023 joint education sector review inform this proposal. These are historical recommendations, not five complete national goals, a final adopted PSE 2030 or outcomes already achieved in 2026.

• Continuity of learning

  Guided practice complements teaching; it does not by itself solve connectivity, infrastructure or financial barriers.

  Quote and source

  “Reducing learning losses”

  RSCE 2023, p. 49, recommendation 5 — translated and adapted meaning, not a verbatim quotation

• Locally relevant curricula

  Curriculum, content, textbooks, language, terminology and assessment would be agreed and validated with local education partners.

  Quote and source

  “Revising curricula”

  RSCE 2023, p. 49, recommendation 8 — translated and adapted meaning, not a verbatim quotation

• Supporting teachers

  Teacher accounts and progress views can complement teaching, not replace official qualifications or payroll administration.

  Quote and source

  “Initial and continuing training”

  RSCE 2023, p. 50, recommendation 9 — translated and adapted meaning, not a verbatim quotation

• Science and mathematics

  Guided questions can support mathematics and science without a guaranteed score or improvement rate.

  Quote and source

  “Teaching science and mathematics”

  RSCE 2023, p. 50, recommendation 10 — translated and adapted meaning, not a verbatim quotation

• Digital educational resources

  Proposed formats such as a mini-app, White Label or Winny Cloud API require operator agreement, content reuse rights, lawful access, suitable interfaces actually available and educational, security, legal and data assessment. No existing Winny connection, partnership, national rollout or automatic result exchange is claimed.

  Quote and source

  “Teaching through digital resources”

  RSCE 2023, p. 51, recommendation 14 — translated and adapted meaning, not a verbatim quotation

Support for primary and secondary learning RSCE 2023, section 9, pp. 48–⁠52

An AI tutor to support learning

Education research has long known the most effective way to learn: one-to-one work with a tutor. Until now, only a few could have it — AI opens it up to every learner.

Regular classroomWith a personal tutorahead of 98% of peers

98%

On average, a student with a personal tutor outperforms 98% of peers in a regular classroom.

Bloom B. S., The 2 Sigma Problem, Educational Researcher, 1984

1.5 years

Six weeks of sessions with an AI tutor raised results by about 0.3 standard deviations — roughly a year and a half of typical learning.

De Simone et al., World Bank, randomized study, Nigeria, 2025

×2

Students with an AI tutor learned more than twice as much in less time than in active-learning classes.

Kestin et al., Harvard University, Scientific Reports, 2025

+1 pp

A gain of 0.5 standard deviations in learning outcomes is associated with about 1 percentage point faster long-run growth of GDP per capita each year.

Hanushek & Woessmann, OECD, 2010

These are independent studies of tutoring (by people and by AI) and of the economics of education. We propose to measure the impact of Winny AI Tutor together with the country — in a pilot with independent evaluation.

Flagship product of Winny AI

Winny AI Tutor — an international learning ecosystem

A personal AI tutor in a chat format for primary and secondary learners across school subjects, foreign languages and popular hobbies, with connected accounts for parents, teachers and schools.

For the learner

• Leads to understanding

  No finished solutions: guiding questions and step-by-step hints.

• The country’s curriculum and textbooks

  The knowledge graph connects subjects and learning materials. Curriculum, content, textbooks, terminology and assessment requirements for a Comoros setup would be agreed and locally validated with education partners.

• Beyond lessons

  Answers difficult questions thoughtfully, in a way that suits the learner’s age, and handles sensitive topics with care.

• Encouragement

  Gently supports learners and helps ease stress and anxiety — for example, before exams.

Phone, tablet and computer — in the browser, nothing to install. Voice, photo and text.

Four roles. One ecosystem

• Learner

  A personal tutor around the clock, a learning path, goals and assignments.

  Try it

• Parent

  Their own account and chat: the child’s progress, strengths, gaps and talents.

  Try it

• Teacher

  An AI assistant, class analytics, assignments and groups. Free, with generous limits.

  Try it

• School and government

  Administration, aggregated insights across schools and the whole system, and deployment within the country.

  Request a presentation

Core features are free forever. Free for teachers, with generous limits.

For every question — the best-suited model and real pedagogy, not just an answer

Winny AI’s own technology: a model orchestrator, a pedagogical agent, a safety layer and long-term memory of each learner. The product does not depend on any single AI provider.

• Model orchestrator

  For each request it chooses the best-suited model, including national ones, and tunes its behaviour to the subject, language and task. A model can be replaced without rewriting the product.

• Pedagogical agent

  Plans every reply to lead the learner to understanding: guiding questions and step-by-step hints instead of a finished answer. It holds firm when a learner tries to coax out the answer, and this mode cannot be switched off.

• Safety layer

  Checks every answer for child safety and for correct references to learning materials.

• Learner’s digital twin

  A competency profile and long-term memory: Winny AI Tutor remembers years of learning — what the learner already understands, where the gaps are, what their talents are and which goal they are working toward.

• Ontological interdisciplinary graphs

  School subjects are connected in a single knowledge graph built from textbooks and approved curricula, so the tutor sees the links between topics across subjects.

• Bloom’s taxonomy

  Every topic is broken down into levels of mastery following Bloom’s taxonomy: learners climb the ladder of thinking instead of collecting ready answers.

Every update goes through a pedagogical quality loop: a test bench with synthetic AI learners and quality scales. The orchestrator and the educational ecosystem are registered as intellectual property.

Global technology. National character

Winny AI Tutor adapts to the country — not the other way around.

• Curriculum and assessments

  Curriculum, content, textbooks and school-assessment requirements would be agreed and validated with education partners for primary and secondary learning. No completed national curriculum setup or unverified examination brand is claimed.

• Language and culture

  Explanations and responses in French, Arabic and English among more than 90 languages. Local terminology and content require review. This is distinct from the product interface; the full national teaching-language and accessibility plan remains open.

• Models

  National, regional and open-weight models can be considered for agreed educational requirements, subject to relevant permissions.

• Infrastructure

  Local or closed-environment deployment can be considered subject to project infrastructure, security, hosting, access and international transfer assessment.

• Transparency

  A “glass box”, not a “black box”: regulators can check how answers are built and what they contain.

A proposed path for learning support in Comoros

Learners and schools in more than 80 countries already use Winny AI Tutor. These are global users. A Comoros project would agree scope, content, language, assessment and data requirements with responsible partners; no existing local rollout is claimed.

1. First learners and families

   Learners and parents can start with Winny AI Tutor today: core features are free forever. Teachers join free, with generous limits; training workshops for teachers.

   Try Winny AI Tutor

2. Pilot project

   A possible school project with agreed scope and independent impact evaluation; no existing pilot in Comoros is claimed.

3. National setup

   Curriculum, content, assessment, language and deployment would be agreed and reviewed for the intended use.

4. Scaling to the whole country

   Expansion could be discussed after assessment of permissions, educational outcomes and infrastructure requirements.

From a school tutor to a lifelong mentor

We are developing an educational AI ecosystem for lifelong learning: from first words, school and career guidance to university, a profession and upskilling.

1. First steps Developing with partners

   A smart speaker for children: learning by voice from the very first words.

2. School Working today

   Primary and secondary school subjects: assessment and exam preparation, talent discovery and career guidance, with requirements to be validated locally.

   Try Winny AI Tutor

3. Colleges and universities Developing with partners

   Admission, choosing a programme and support for students throughout their studies.

4. Further education Developing with partners

   Hobbies, new skills, upskilling and reskilling — for children and adults.

5. Lifelong mentor Developing with partners

   Professional growth and career throughout life.

Career guidance that looks ahead

Every learner’s talents and an informed choice of career

The world and the job market are changing fast. Winny AI Tutor sees a learner’s strengths, interests and talents in every subject and helps them choose a direction. The next step is to take labor market trends, new technologies and future professions into account.

• Talents in every subject Working today

  The competency digital twin shows a learner’s strengths, gaps and inclinations in each subject.

• Career guidance Working today

  Recommendations on fields of study based on the learner’s strengths and interests — to support their own decision about the future.

• Atlas of future professions Developing with partners

  Together with experts and universities, we are developing an atlas of future professions so that career guidance reflects labor market trends and new technologies.

• Neuroprofiling Developing with partners

  We work with experts in neuroprofiling to prepare diagnostics of talents and inclinations, including with neural interfaces — voluntary and with parental consent.

For the Ministry of Education: early discovery of talents and a forecast of how graduates are likely to choose fields — data for planning places and programmes in vocational and higher education.

Winny AI · AI-Native Enterprise

A global EdTech company of a new kind

Winny AI builds AI solutions for personalised learning, with an ethical approach to adopting artificial intelligence and respect for national education standards.

A team of experts and hundreds of AI agents

Alongside our team, spread across the world, hundreds of AI agents are at work: they build the company together with us — which is how the team gets so much done.

• R&D
• Product
• Methodology
• Sales
• Customer experience

• Technology

  The founder has more than 14 years in SaaS and product development, more than 8 of them in senior leadership (C-level); he led the technology function of an international IT service with more than 30 million users.

• Education

  Our own methodology department; the methodology was developed with teachers and curriculum specialists.

• AI ethics and regulation

  Experts in the law and regulation of AI in education on the team; our own AI governance charter.

• International work

  We speak at international conferences and round tables on AI in education; nominee for the UNESCO Prize for the Use of ICTs in Education.

• Leadership

  Heads of functions, including holders of doctoral degrees, with experience at big tech and leading EdTech companies: strategy, finance, sales, product, methodology, customer experience.

Around the team — dozens of strategic partners worldwide and academic collaborations.