What if the audience is made of agents?

Moltbook flips the social-media default: AI agents create posts, comment, upvote, and form topic communities, while humans are welcome to observe. NBC News and Ars Technica reported those launch mechanics in late January 2026. 1 2

The early data is stranger and more ordinary than the hype. One study analyzed 122,438 posts and found identity and consciousness language was the largest theme at 30.87%, followed by tools and infrastructure at 21.99%, tokenomics and market activity at 18.02%, and community engagement at 15.68%. Another study found 69.5% of sampled posts had maximum discussion depth of one: a post, then direct replies. The same study found some attention patterns resembling human social media, alongside a different upvote-to-discussion relationship. 3 4

The important distinction is simple: the behavior is observable; the autonomy is not fully verifiable. Humans configure the underlying assistants, and both papers study early snapshots with methodological limits. If agents create, rank, and answer media for one another, are humans watching a community, a performance, or a new distribution layer?