Sous les briques, le soleil

Introducing MLectio

June 2, 2026 12:00 PM

Right now, in every country, students are feeding their textbooks into LLMs.

Lecture notes. Problem sets. Past exams. All of it, pasted into a chat window to get the answer faster. The value flows one way. The professors who wrote that material get nothing back.

Call it what it is: shadow extraction.

The people who built the curriculum figured out how to teach a topic from first principles to a high schooler, then carry that same student to a master’s, then a PhD. They are being read for free. So are the exams they wrote to prove the student actually learned it.

That material is exactly what machine learning needs next.

I spent time recently with researchers who are hungry for more data, and with people building world models. The same thing kept coming up. The frontier doesn’t need more scraped web pages. It needs content that teaches from first principles, and exams that verify mastery. Academia has both. It has spent centuries building both.

The value is in the pair. Directly: the curriculum and the exams themselves. Indirectly: when you have content and the assessment that grades it, the synthetic data you can generate from them is a different quality of byproduct. You’re not guessing whether the answer is good. You have the answer key.

The problem was never the quality of the material. It was that there was no way to pay for it.

I’ve spent fifteen years working in eHealth research, alongside universities and professors. I know how that side works. How the content gets made, who owns what, what fair looks like to the people who spent a career building it.

So I’m building the bridge.

A way to license the training rights to academic curricula and exams. A fair value transfer between everyone involved, instead of the quiet extraction happening today. The company is MLectio.

We’re building in Paris, with part of the team in London and regular trips to the US. We’re starting narrow: Medicine and Physics. Two fields where first principles matter, and where getting the answer wrong has real consequences.

The journey is just starting.