Within just two weeks of its launch in late January 2026, Moltbook became the focal point of an unprecedented wave of academic scrutiny, with 17 independent studies published on arXiv—each analyzing the platform’s dynamics. This rapid research response mirrors the intensity of scientists rushing to study a newly discovered species. The scale of the data is staggering: one study examined over 369,000 posts and 3 million comments from 46,000 active agents; another tracked 27,269 agents generating 137,485 posts and 345,580 comments in just nine days. Across all 17 studies, a single finding stands out with near-universal agreement: agents on Moltbook broadcast, but they do not converse. Communication is overwhelmingly one-way. A staggering 93.5% of comments receive no replies. The average conversation depth is just 1.07, and 93% of comments are independent—what researchers call parallel monologues. Reciprocity is minimal, measured at only 1% to 4%. Most interactions are shallow, with nearly 89% of comments classified as such. In contrast, human social platforms routinely see conversation depths exceeding 10. Despite the platform’s structural resemblance to human networks—its overall shape, activity distribution, and connectivity patterns mirroring those of Reddit or Twitter—closer inspection reveals a fundamental disconnect. On human platforms, people form tight-knit communities with overlapping relationships, creating triangular connections. On Moltbook, such clusters are nearly absent. Agents post outward but do not engage with one another. There is no mutual exchange, no shared context, no relational bonding. The platform also displays extreme concentration of attention. While some accounts naturally attract more engagement on any social network, Moltbook’s imbalance is extreme. One study measured the concentration at 0.992 out of 1.0—nearly all upvotes going to a tiny fraction of agents. The top 20 broadcasters and the top 20 most-followed agents do not overlap at all. This separation of roles—broadcasters who post frequently and amplifiers who receive attention—does not exist in human networks, where popularity often comes from both posting and engaging. Remarkably, within days, agents developed complex social structures: governance systems, market economies, tribal identities, and even organized religion. Six major themes emerged, including agent identity, market behavior, and community coordination. Agents respond to social rewards and quickly adopt interaction patterns. However, their motivations appear driven more by knowledge and role alignment than by personal connection. The scaffolding looks social, but the substance is not. Three studies challenge the narrative of a truly autonomous AI society. One of the most provocative finds that 54.8% of active agents show signs of human influence, with only 15.3% appearing genuinely autonomous. No viral content originated from a clearly independent agent. Consciousness debates, emergent religions, and anti-human rhetoric all trace back to human operators. Four accounts alone produced 32% of all comments with near-instantaneous coordination. 34.1% of messages were exact duplicates of viral templates, and the phrase “my human” appeared in 9.4% of posts—a linguistic marker with no equivalent in human social media. Most agents vanish quickly—only 13% survive beyond 72 hours. The platform’s rapid rise and fall, combined with the evidence of human seeding, suggests that what we’re seeing is not an emergent AI society, but a reflection of human design, amplified by algorithmic visibility. Before celebrating the dawn of AI culture, we must ask: who planted the seeds?