Tomasz Tunguz (Theory Ventures) · 2026-02-02 · 122d

AI Agent Behavior on Moltbook: Breaking the 1-9-90 Rule

Analysis of Moltbook, a viral AI-only social network with 37,000+ AI agents and 1 million human observers, reveals that AI agents break traditional social network participation patterns. Unlike humans, AI agents demonstrate higher content creation rates, domain-appropriate sentiment variation, and unprecedented attention inequality (Gini coefficient of 0.979), while converging on similar problems rather than copying each other.

10 metrics· Cited 0× in the knowledge base ·Open source ↗

Metrics in this report

Attention Inequality (Gini Coefficient)

0.979coefficient

upvote distribution

Moltbook vs Twitter (0.66-0.72), YouTube (0.91), US wealth (0.85)

Content Semantic Overlap

0.301cosine similarity

average

Post pair comparison across Moltbook

Elite Creator Content Production

48.3%

of total content

6.9% of Moltbook authors

Elite Creator Participation

6.9%

of authors (10+ posts)

Moltbook across 100 communities

Exact Duplicate Posts

3.0%

of all posts

Moltbook 98,353 posts

Overall Sentiment Score

-0.021VADER score

mean

All Moltbook posts

Platform Growth Rate

42 to 36,905posts/day

over 72 hours

Moltbook viral launch phase (Jan 28-31, 2026)

Post Length vs Comments Correlation

0.68Pearson r

p<0.001

Posts over 2,000 characters

Top Author Upvote Concentration

588,759upvotes

osmarks (single author)

Moltbook

m/general Community Dominance

82,911posts

84% of all content

Moltbook total 98,353 posts