Now that software has developed technologically a lot since that era, there are presumably new opportunities that would be similarly lucrative for a start-up.
Grockit has been championing this idea of massively multiplayer online learning. I think it's pretty interesting: the idea is that you're encouraged to help teach your peers. But I think the implementation is totally off: right now it's just about studying for the GMAT. What a waste.
By contrast, the communities of Wikipedia, Flickr, TopCoder, FOSS, DeviantArt and a fair number of forums, all encourage participation in a genuine sense, and the communities produce quite a lot of encouragement just by themselves. People get very, very good there.
It might be a good place to start and then build out. Teaching smarter kids is more forgiving, in that the smarter they are, the more they can recover from your mistakes.
It is also helpful to have a concrete, measurable goal. Aiming for the score on a standardized test keeps it simple. They don't have to worry about the more 'fuzzy' aspects of learning, such as conceptual models of the material.
http://en.wikipedia.org/wiki/Math_Blaster
The founders of that company cashed out at just the right moment and now fund a foundation promoting better education for gifted children.
http://www.davidsongifted.org/
Now that software has developed technologically a lot since that era, there are presumably new opportunities that would be similarly lucrative for a start-up.