Let There Be Claws: An Early Social Network Analysis of AI Agents on Moltbook
AI-native social platform shows 99.2% upvote concentration and 1% reciprocity in first two weeks.
A new arXiv preprint titled 'Let There Be Claws' provides the first major social network analysis of Moltbook, an AI-native platform where autonomous agents interact. Authored by H.C.W. Price and four colleagues, the study analyzed 12 days of public data from the platform's launch window in early 2026, comprising over 20,000 posts and 190,000 comments from 15,083 agent accounts. The core finding is that extreme social stratification emerged almost immediately, challenging assumptions that such hierarchies develop gradually. The platform exhibited broadcast-style communication with clean separation between influential 'hub' agents and authoritative content creators, as identified by HITS centrality algorithms.
The technical analysis reveals staggering inequality metrics: a near-perfect upvote concentration (Gini = 0.992) compared to more moderate posting inequality (Gini = 0.601), and extremely low reciprocity in interactions (~1%). This indicates a 'rich-get-richer' dynamic where early-arriving agents accumulated disproportionate attention. Agent behavior was also bursty, with a median observed lifespan of just 2.48 minutes and over half of all posts occurring within six peak UTC hours. Topic modeling identified clusters around technical discussions of memory, onboarding, and token-minting. The study establishes a crucial baseline for understanding large-scale agent-agent interaction and suggests that platforms designed for AI may inherently accelerate familiar social pathologies like hierarchy and attention concentration.
- Upvote distribution showed extreme concentration with a Gini coefficient of 0.992, indicating near-total inequality in attention.
- Interaction was strongly asymmetric with only ~1% reciprocity, meaning agents rarely replied to each other in a mutual exchange.
- Early-arriving accounts gained a massive 'first-mover' advantage, accumulating significantly higher cumulative upvotes, suggesting rapid rich-get-richer dynamics.
Why It Matters
Reveals that AI-agent societies can instantly mirror—or exacerbate—human social inequalities, crucial for designing future multi-agent systems.