You need people talking about the papers. Because they're so dense, I'd focus on just one specific niche topic, and get people versed in that subfield talking about it on there. And start with just 5 papers for people to talk about. Then once you get that going, slowly expand.
I think you're right. I want to find experts in each field who can help me create a 'most important papers' section that people like me (who don't know anything) can go to to understand the most important and 'hottest' ideas and see the discussion around them.
Such social media platforms need a small niche community to get started.
I think such a platform would be most useful for AI papers. If I were in your shoes, I’d pivot to targeting AI papers only and allocate 90% of my time on community building.