The dream!
I've seen some papers on some experimental systems that can work in a peer-to-peer environment, that try to balance our privacy (since somehow we need to collaborate without exposing all the content we like or are interested in) and good recommendations. It is however quite an endeavor implementing such systems.
To add to this, most platforms, at least in Europe, but I guess its the same everywhere, are required by law to give you all the data they have on you. By building "adapters" for the major platforms, in order to process their formats, you wouldn't even need to start from 0 when "migrating" to the open source socials. If rendered automatic, even the average person could use it.
Btw I as well I am starved for time, otherwise I would love to try my hand at it, but the best I could do right now it's just maybe help. I am a Machine Learning dev, so I have some practical knowledge in that domain at least, although not on reccomender systems :/.
Thank you for the input! Seems like a step in the right direction at least, but I would like something more peer-to-peer than federated, so that I myself can choose and direct the algorithm and not the server I happen to use as an entry point. I'm guessing, though, that this might not exist at this point.