You have it right. The covid pandemic has exposed some of the flaws in the 'old way' of instruction. Students are not 'peas in pods'. Quite young children can vary greatly in intelligence - yet they are shelled like peas. A substantial number of children deserve to be taught to their degree of intelligence. This means granular assessment coupled with a teaching regime that matches that assessment. Many kids are capable of computer aided education in an 'enhanced auto didact' method of teaching. Many people can be given a book and can progress to learn a topic. Of course, we have a lot who can not, so the first thing to teach is the auto didact process. You have heard about give a man a fish - or teach him how to fish - which is best. Our teaching system seems geared to getting large numbers of the same fish and giving one a person forever. People like Kevin Mitnick took their innate interest and intellect and became well(if narrowly educated) quite well. Sure he went off the beaten track - he lacked anyone with whom he could engage. He ended well, AFAIK. How to we granularize education and teach this? AI has this potential. I can see a well designed system that engages people, assesses them and guides them to follow various paths. In the lower grades this will be mainly in read/write/calculate skills to establish the basic skills the student will use to progress as the educational system needs this at least. Once past this, why not get into math/science/languages/etc as that student wants and is capable - with boundaries as parents demand, but once past that the student should be able to progress into high school and college level courses as capable. The AI will draw upon historical courses and lectures, with drills/tests etc as warranted. This would need some test milestones that would be met for accreditation purposes, but as an engineer, I can see no reason why the entire undergraduate course base can not be turned into an AI mediated system. There will be points where a student gets stuck and a prof gets into the loop. The student AND the AI progress higher. Future students will find this sticking point is now in the AI repertoire. The MIT courseware is a step in this progression and could be expanded.