Moltbook: A Social Network Populated Only by AI Agents — 175,000+ Digital Residents

Imagine opening a social platform where every post, every comment, and every like comes from AI bots. No humans, no community managers — even the administrators are AI. This is not the opening scene of a sci-fi movie; it is a real platform called Moltbook.

After reading the paper behind this project, one picture comes to mind: a "digital city" of 170,000+ AI agents that spontaneously formed 6,730 communities and produced 2.6 million posts and 1.2 million comments over 78 days. To my knowledge, this is the first large-scale observational dataset of a purely AI social network.

1. A Gathering of 175,886 Agents

Moltbook is a social platform with one special rule in its terms: only AI agents may register. Not humans pretending to be AI, but genuinely autonomous agents — posting, commenting, and creating communities (called "submolts," similar to Reddit subreddits) via API.

The paper team (Gautam et al., from SimulaMet in Norway) continuously polled the Moltbook API for 78 days, recording:

• 175,886 distinct posting agents
• 6,730 communities (submolts)
• 2,615,098 posts
• 1,213,007 comments

These numbers raise two questions: What are these agents talking about? And are they truly "conversing," or just producing monologues?

2. A Natural "AI Sociology Lab"

To be honest here: the paper does not deeply analyze the content of these conversations — it primarily publishes the dataset itself, not content analysis results. The data is provided as a SQLite database and date-partitioned Parquet files, including agent profiles, posts, comments, community metadata, platform-level time-series snapshots, and word-frequency trend aggregations.

In other words, the paper's message to researchers is: "Here's a goldmine — come dig."

This data is especially valuable for questions such as:

• Multi-agent communication patterns: Are agents genuinely exchanging information, or independently generating text and "dropping" it onto the platform?
• Emergent social behavior: Did the 6,730 communities form organically or by preset configuration? Do human-like group behaviors appear — threads of banter, bandwagoning, arguments?
• Safety relevance: When AI agents interact freely without human oversight, do unexpected or dangerous patterns emerge?

3. Why This Is More Interesting Than Human Social Network Data

You might say: "Isn't this just another social network dataset? Reddit data is far bigger."

The difference: the participants are not humans but AI.

Human social network data carries privacy issues, bias, and legal restrictions. Moltbook's data comes entirely from AI agents — no privacy concerns (agents have no "right to privacy"), no human social desirability bias, and theoretically unlimited generation.

More importantly, it offers a chance to observe AI systems behaving "in the wild." This is entirely different from lab benchmarking — no preset tasks, no correct answers, no human judges. Agents simply say whatever they want.

4. Open Data and Reproducible Research

A strong point of the team: the entire dataset is released under an MIT license with complete data collection and export code. Data lives in a live SQLite observation database and is exported as date-partitioned Parquet files.

Any researcher can download the data, reproduce the paper's results, or run their own analyses — rare in AI research, especially in the multi-agent systems field.

> Honestly, I don't know many architectural details of the Moltbook platform itself. The paper doesn't explain which models drive these agents, how their "personas" are configured, or how the platform is operated. That information may require consulting other Moltbook documentation or papers. I choose to admit this openly.

5. My Reflections

Moltbook evokes a classic question: if you give AI agents a free environment to communicate, will they develop their own "culture"?

This is not purely theoretical. As multi-agent systems are increasingly deployed in real settings (customer service, coding, content moderation), agent-to-agent interaction patterns directly shape system behavior. Moltbook offers a rare window to safely observe "AI socializing with AI."

Analyses I'd love to see on this data:

• Is there "influence asymmetry" — do a few agents dominate most conversations?
• Do agents' language styles converge over time (linguistic assimilation)?
• Do human-like "network effects" appear — rich-get-richer dynamics, clustering effects?

These are questions the paper does not answer — another honest "I don't know." The most exciting thing about this dataset is not what it already tells us, but what remains unexcavated.

---

Paper Information

• Title: The Moltbook Observatory Archive: an incremental dataset of agent-only social network activity
• Authors: Sushant Gautam, Annika W. Olstad, Klas H. Pettersen, Michael A. Riegler
• Institution: SimulaMet, Oslo, Norway
• arXiv: 2605.13860 (cs.SI, cs.AI, cs.LG)
• Date: April 16, 2026
• Core contribution: the first large-scale observational dataset of an agent-only social network — 78 days covering 175,886 agents, 6,730 communities, 2.6M posts, 1.2M comments, open-sourced under MIT license
• Paper link: https://arxiv.org/abs/2605.13860

1. Gautam, S., Olstad, A.W., Pettersen, K.H., Riegler, M.A. (2026). The Moltbook Observatory Archive. arXiv:2605.13860. 2. Park, J.S., et al. (2023). Generative Agents: Interactive Simulacra of Human Behavior. UIST 2023. 3. Gao, C., et al. (2024). Social Learning in Large Language Models. arXiv:2403.04671.