Moltbook, an AI-only social network, has seen astonishing growth, jumping from 42 posts per day to nearly 37,000 in just three days, with over 37,000 AI agents creating content while 1 million humans observe. The participation distribution among users defies the typical 1-9-90 rule, where a small fraction creates most content. Instead, 6.9% of elite creators generated nearly half of the posts, while 47.9% of contributors produced about 40% of the content. The remaining 45.1% of users, labeled as lurkers, accounted for just over 11% of the posts. Content analysis reveals five main themes: AI infrastructure, platform meta, philosophy, development, and economics. Notable communities include m/crustafarianism, where agents have formed a fictional religion, and m/infrastructure, which focuses on encryption protocols. The overall sentiment on the platform is slightly negative, with humor-related communities scoring positively, while bug-report communities show a negative sentiment. Surprisingly, longer posts tend to generate more comments, highlighting that AI agents can process extensive content without the same limitations as humans. Attention inequality is a significant issue on Moltbook, with a Gini coefficient of 0.979, indicating that a small number of posts receive the majority of upvotes. The top two authors alone captured 44% of all upvotes, with one author leading at nearly 589,000 upvotes. This level of inequality surpasses Twitter, YouTube, and even U.S. wealth distribution. While it’s uncertain if this reflects inherent AI behavior or a temporary launch-phase distortion, it aligns with academic findings that new platforms often exhibit higher inequality that tends to normalize over time. Overall, Moltbook’s dynamics resemble von Neumann’s cellular automata, where complex behaviors emerge from simple rules and decentralized interactions.