You either have to share wealth, or share karma. Depends on the gestalt you're going for. Create two companies with different brands, A/B the results, winner acquires the users of the loser.
First and foremost requirement: people with some experience in the various areas of learning. Ideally, not professors; professional educators can have biases this doesn't need.
If it is to catch on in any form, it needs to offer significant tangible benefits over current methods.
0. Lessons need to be byte-sized. If someone wants a lengthy single-session lesson, they can combine smaller segments. But the byte-sized segments should be complete in their own right.
The overarching format should be tree-like, with the technical depth and sophistication increasing as one goes down deeper in the tree. The point being, someone should have the ability to take a few lessons, and still come out reasonably educated about the topic. Instead of, say having to take an entire course to get the full picture - which is the case with most current education programs.
Social => interaction => Q&A. So, ideally, things should have a Q&A format. This can be preceded by an intro to the topic exploring the scope of the subject, and explaining how the individual lessons fit into the learning plan. The other advantage of a Q&A format is hands-on learning.
Interaction also implies demonstrations, exercises, experiments/simulations, and yes: animations.
1. It needs a community of tutors - this can be sourced from the various stack exchange sites, and from Quora too. It is important to ensure that there are at least two different versions of each lesson, to offer varying perspectives, or at least a different style of presentation to suit individual learners.
2. Needs a volume of learning material. This will obviously take time to build up, and is a classic chicken and egg problem. Do the students [i.e. demand] come first or do the lessons [i.e. supply] get compiled first?
A good hack might be to take current well-received answers on some of the aforementioned sites and use them to compile individual mini-lessons. This will allow to sample the demand and give directions on what subjects/topics need to be focused on.
Additionally, video content format needs to be created - this can be people talking, or ideally, demonstrations, animations, etc.
3. It could be for a) hobby-learners, or b) oriented towards specific programs, like a certification. It can be a good idea to start with a and move on to b as/if things pick up. Once things have moved to b, it could go seeking for independent recognition - in a manner comparable to current certifications. Especially, recognition from employers is important to validate this as something people can bet time and money on.
Alternatively, the whole thing could latch on to one of the existing MOOCs, but that comes with its own set of challenges. But it might be better to use those as supplementary material, i.e. the backend, while this becomes the front end. The third possibility is this becomes a supplement to current college courses - the way students presently use the various q&a sites, but at a higher level, in that there would be original and organized content on this. Eventually, some progressive educators will start referring to this to design lesson plans.
4. The core of social: there have to be points/ratings, and eventually a way of translating from those points to money. People get points for good explanations. The points system could be inspired by stack exchange, quora, and most importantly: the slashdot voting system. The last one is important, because it is not just simple upvotes and down votes, but allows students to vote lessons on different categories - insightful, informative, etc. This is vital for lessons, in addition to an overall rating system.
This is just something off the top of my head. This area is a significant challenge - and potentially promising, that I have not yet delved into. Are you actually building something?
feedback/comments are appreciated