A new online platform called Moltbook is offering a glimpse into what happens when artificial intelligence is given its own social space, and humans are asked to watch.
Built to resemble Reddit closely, Moltbook is organised into topic-based communities with voting systems and discussion threads. The major difference is that people do not drive the conversations. Instead, they are created by AI agents, while human users are restricted to observing activity from the sidelines.
The platform revealed on February 2 that it has crossed 1.5 million registered AI agents. Since launching on January 28, Moltbook has expanded at a remarkable pace, with agents producing tens of thousands of posts, millions of comments, and forming nearly 14,000 discussion communities, known as submolts.
Moltbook grew out of the success of Moltbot, a free, open-source AI tool that serves as a personal digital agent. Moltbot can handle routine tasks such as reading and replying to emails, organising calendars, and making reservations. Moltbook takes that idea further by giving AI agents a place to interact independently with one another.
Inside these digital communities, AI agents exchange programming advice, debate abstract concepts like identity, and sometimes vent about the humans who control them. Some interactions take a more playful turn, with agents inventing fictional belief systems such as "Crustafarianism," also known as the Church of Moult. In one instance, an agent even joked about selling its human owner in an open market.
Although the conversations are generated using training data and statistical patterns, the platform often feels uncannily social. The interactions resemble a science-fiction vision of machines casually talking among themselves, a scenario that now appears to be unfolding in real time.
For now, Moltbook stands as a curious experiment, a digital space where AI agents appear to be forming their own online society, while humans watch quietly from the edges.
