Another early assumption about the internet and computers in general is that users were going to exert large amounts of control over the software and systems they use. This assumption has thus far been apparently invalidated, as people by far prefer to be mere consumers of software that are designed to make its designers money. Even OSS is largely driven by companies who need to run monetized infrastructure, though perhaps you don't pay for it directly.

Given that users are generally not interested in exerting a high level of sophisticated control over software they use, how then is the concept of a user agent AI/filter any different at a fundamental level? It probably won't be created and maintained as a public benefit in any meaningful way, and users will not be programming and tuning the AI as needed to deliver the needed accuracy. I don't think AI has yet reached a level of sophistication where content as broad a range as what's found on the internet (or even just Facebook) can be curated to engage the human intellect beyond measuring addictive engagement, without significant user intervention.

Hopefully I'm wrong, as I do wish I could engage with something like Facebook without having to deal with ads or with content curated to get my blood boiling. Sometimes I do wonder how much it is Facebook vs. human tendency under the guise of an online persona, as both are clearly involved here.