In the legacy education model, teachers combine coaching, feedback, and content delivery. By coaching I mean advice, guidance, and encouragement. Feedback includes formal grading as well as informal praise and criticism. Content delivery includes lectures and reading assignments.
Perhaps the key to radically changing education is to break up those functions.
1. The coach should be someone who knows the student well, who can relate to and motivate the student, who can recommend a good educational path, who takes account of the student's strengths and weaknesses, and who stays on top of how well the student is doing relative to the student's ability.
2. The formal feedback can come from strangers. Students can solve problems or write essays and have these graded by a separate service.
3. The content delivery should be "pulled" by the student rather than pushed by a teacher. For example, a student and a coach could agree that the student should learn statistics. The student then selects a statistics curriculum and works through it. The Khan Academy lectures on statistics are particularly good, in my opinion. But Carnegie-Mellon has a good on-line stats course, also. My guess is that, overall, there is enough content on line to obtain a world class education.
4. In addition to teachers, students can learn from peers. I think that peer learning works best when students are close to one another in terms of ability and motivation. My hypothesis is that ten students who are similar in ability and motivation but separated by distance can learn better together than ten students who are in the same room but with disparate levels of ability and motivation in the subject.
This is pretty close to the way I learned economics. Bernie Saffran was mostly a coach, giving us suggestions for topics to study. Ultimately, we were assessed by strangers (Swarthmore College has an Honors Program under which outside examiners come and administer tests).
Most education reformers want to focus on low-end students. While this is a noble idea, I think it is not a good path for reform. When you fail, you do not know whether it is because the innovations were not good or because the student population is too difficult to reach.
If I were trying to implement a school redesigned along the lines I describe, I would start with high-end students. Take some young people aged 16-18 who are bright misfits. The types who already would rather read on their own than sit in class. Offer them the alternative system described above. If it fails for them, you know you've got a bad model. But if you get it working for them, then you can try to spread it to other types of students.