The problem that we are facing is one of parasocial asymmetric relationships caused by the power laws inherent in ad funded, engagement based social media, optimised for time-on-device.

Everyone comments as though they know Peter (and other public figures), and that feeling makes them frustrated because they'll almost never respond to them specifically. That frustration piggybacks on what they want to say, making it aggressive and unreasonable. Nobody would say these things to strangers in public, or if there were a real prospect of a response.

Add to that the fact that transgression harvests engagement and a share of the platform's ad revenue and we have a competitive discourse that is now a race to the bottom.

The only chance of turning back the tide of degrading public discourse is to aggregate real world event data using LLMs and build a social platform entirely oriented towards bringing people together in physical spaces, funded like Substack and Patreon but for real world creators (artists, speakers, teachers, musicians, DJs, intellectuals, etc), using the same perk system for accessing real world types of exclusivity. LLMs can do this because they solve the economic problem of efficiently aggregating event data and tracking changes, because they can structure unstructured data.

It's then possible to build a social network optimised for consensus rather than engagement, underpinned by the Pagerank algorithm.