So far Tor has implemented WebRTC[0], but that's just for bridging to their main network.
[0] https://snowflake.torproject.org/
P.S. I've also considered the "decentralized p2p social media" idea myself, but mostly because I believe the ideas we have currently behind online voting, ranking and moderation are completely at odds with IRL discussions which are P2P and based on "forwarding" ideas to known peers (friends, family, community members, countrymen) rather than posting and ranking content with anonymous peers (which are susceptible to Sybil attacks). The fact that so much discussion takes place on corporate-owned forums (Including this one, regardless of how benevolent ycombinator may be) presents a major threat to democracy in general.
Instead of "liking" or "upvoting" a post on a centralized forum, why not "rehost" or "forward" a post on a decentralized forum: essentially seeding it like in BitTorrent or "pinning" it in IPFS. "Followers" of a user donate their storage and bandwidth to them, combating bureaucratic attacks like delisting and DDoS against popular users.
If you could port IPFS to run completely in a browser you would have this complete "pseudo-social-media" functionality. They have something called "IPNS" where instead of giving someone the hash of a file like in BitTorrent's DHT, you could give them a public key which you use to sign the latest version of a file that is to be fetched. The Public/Private keypair could represent a user's identity, and the file in this case could be a blog or account page which is updated with new links to the user's posts, or links to other user's posts.
So if you ported IPFS to work within a web browser, it would just be a matter of implementing a user-interface. Boom, social media solved. You could maintain parity between desktop and web versions by using libraries like libdatachannel[1] and datachannel-wasm[2]
[1] https://github.com/paullouisageneau/libdatachannel [2] https://github.com/paullouisageneau/datachannel-wasm
I built a small proof of concept along these lines a while back. The implementation was similar to what you're describing with peers seeding and forwarding messages. I didn't get as far as name resolution since that's a tough nut to crack in p2p but building off of IPNS would probably be the way to go. The prototype worked but it wasn't very attractive to use since it was only good for passing notes and there are better services out there with more features. I think something like it would only be worthwhile building if there really was a concerted effort to lock down all social media in a fairly heavy handed way.
