That's right.

um so in a nutshell harvard ran a randomized control trial it was in a physics department so um the professors did a sort of in-classroom active learning so that was already sort of like the gold standard kind of face-to-face um in-person learning and then the same professors created essentially a kind of custom AI tutor.

So it was built by the instructors themselves and following kind of best practice and the content from the syllabus.

Every student experienced both conditions.

So in week one, half the class got the AI lesson and half didn't, but vice versa.

Everyone sat a pre-test before the lesson and a post-test after.

So that's really where they can, I suppose, getting the information from.

You know,