1) Copy paste of images. This is key. With my mac I screenshot a slide from an anatomy atlas to the clipboard. Then use the image occlusion editor to hide labels on the slide, and then I use the editor to highlight a certain area on the slide and ask "what is this muscle?" 70% of my cards require copy/paste of images. Without this feature I could never switch.
2) Cloud storage that synchronizes flashcards on my mac and mobile device. I do a lot of flashcard review in the gym, bathroom, waiting in line, etc. Without mobile I could not switch.
3) Searchable index of cards. I make a flashcard for every major piece of information that I need to know for school, and I use search to connect that information. For instance we spent the last 6 weeks learning about metabolic pathways and what factors regulate enzymes within them, but we never focused on specific regulators. With Anki, however, I can search for a certain activation enzyme, say protein kinase A, and then have a deck of flashcards showing me what pathways are affected by said enzyme.
Those are the biggest must have features for me that set Anki apart. There are loads of great minimal and social flashcard applications (eg Study Blue or Quizlet) available for average users, eg, someone learning spanish. You'll have to find a way to stand out from them.
There's no excuse for you to use Leitner though. Anki is based on the SM-2 algorithm which is well-documented (though only a starting point).
How accurate can your spaced repition algorithm predict my forgetting curve? The time between repitions can be substantially different depending on the answer I give to Anki (a day between a week or something like that)
In which other aspects do you want to be like Rdio? Closed source, analyzing user data, advertisement, monetization?
Anki does have a huge advantage from its excellent spacing algorithm. The prototype I have working so far uses the Leitner system, which is very basic, but an advantage of keeping all the data on the web is the huge potential for A/B testing to figure out the actual best spacing algorithm.
Lastly, I'm not too interested in generating revenue from this project. I don't even have any ideas for monetizing it. This is mainly to scratch my own itch. If I can make enough to cover the cost of running the project (which is currently zero; it's being hosted on a free heroku dyno), I'll be happy.
It's exactly what I did a year ago when I started work on http://www.memorific.com to build a SRS that was fine-tuned for software developers. We just changed to public beta a few weeks ago and started generating revenue and it's been a great experience helping people learn, grow, and meet their goals for learning.
Keep up the good work on noot.io and have a blast doing it! There's nothing like making a product that people use and enjoy.