You need to think of the graph more like 'the part of the web you can possibly see' and then about how in an open web that means you need some process for reducing the "firehose" down to something drinkable every time you are viewing it.
In the contemporary web, this context is provided by your apps, app settings, and central moderation (for both legality and "the algorithm").
In a p2p web, the users can provide it through context tagging, but it needs to be done across both peers AND content, like URLs, events, locations, etc.
A "territory" here in StackerNews, or a "subreddit" is no different than a tag in the semantic social graph from one key to a post. The tags already are the topics. The tags can also be thread topics, or, you can simply repurpose the primitive of a 'headline' or title for an article with a forum topic. How an app presents the data does not actually require many different data.
Similarly, you can then apply distance filtering across any topic or set of topics, with you at the center, and include the entire firehose if you really want to. There is not a limit to your ability to discover new info, there is only a limit to what any single indexer could include in their graph, same as with Google, or X, etc. But here, the graph data is owned by the users and assembled by the indexers.
Also, don't forget that most people are always connected to many networks, with many sources of new info. Echo chambers only exist for people that create them intentionally, they aren't a phenomenon of tragedy of the commons, or bad design, or something.
You need to think of the graph more like 'the part of the web you can possibly see' and then about how in an open web that means you need some process for reducing the "firehose" down to something drinkable every time you are viewing it.
In the contemporary web, this context is provided by your apps, app settings, and central moderation (for both legality and "the algorithm").
In a p2p web, the users can provide it through context tagging, but it needs to be done across both peers AND content, like URLs, events, locations, etc.
A "territory" here in StackerNews, or a "subreddit" is no different than a tag in the semantic social graph from one key to a post. The tags already are the topics. The tags can also be thread topics, or, you can simply repurpose the primitive of a 'headline' or title for an article with a forum topic. How an app presents the data does not actually require many different data.
Similarly, you can then apply distance filtering across any topic or set of topics, with you at the center, and include the entire firehose if you really want to. There is not a limit to your ability to discover new info, there is only a limit to what any single indexer could include in their graph, same as with Google, or X, etc. But here, the graph data is owned by the users and assembled by the indexers.
Also, don't forget that most people are always connected to many networks, with many sources of new info. Echo chambers only exist for people that create them intentionally, they aren't a phenomenon of tragedy of the commons, or bad design, or something.