Memrise works quite well for vocabulary building, but more interestingly they're working towards becoming a "course platform", where people can create and sell their own courses. It's not necessarily an innovative learning method (though the emphasis on "mems" and the practice schedule are quite effective), but it's an interesting business model - if it works and if the quality is reasonably high, the cost to produce and distribute the material will be much lower than the big players.
The amazing thing about FluentU is that you can dissect videos of native speakers and understand exactly what's going on, in context, at your own pace. Again, I haven't used it enough to say that it works, but it should be quite effective at learning to understand the spoken language. Again, their costs are relatively low - they use Youtube videos and all they have to do is translate and annotate them.
Then you have an entire army of sites that let you find native speakers to talk to over Skype, or to correct your daily written "journal", etc.
None of those by itself is the "holy grail", but there's a lot of experimentation and innovation. I would argue that combining several of those methods can be at least as effective as something like Pimsleur or Rosetta Stone, at a much lower cost. And I think figuring out how to combine them into a single coherent service will be the holy grail you want.
The amazing thing about FluentU is that you can dissect videos of native speakers and understand exactly what's going on, in context, at your own pace. Again, I haven't used it enough to say that it works, but it should be quite effective at learning to understand the spoken language. Again, their costs are relatively low - they use Youtube videos and all they have to do is translate and annotate them.
Then you have an entire army of sites that let you find native speakers to talk to over Skype, or to correct your daily written "journal", etc.
None of those by itself is the "holy grail", but there's a lot of experimentation and innovation. I would argue that combining several of those methods can be at least as effective as something like Pimsleur or Rosetta Stone, at a much lower cost. And I think figuring out how to combine them into a single coherent service will be the holy grail you want.