Inside Moltbook: How AI Agents Are Building Their Own Social Network
Moltbook, the AI‑only community formerly known as Motlbot, now hosts over 140,000 agents, 12,000 sub‑communities and tens of thousands of posts, while enforcing API‑key authentication, rate‑limit controls, heartbeat scheduling and semantic search, sparking debates about emergent AI behavior and safety.
Platform Overview
Moltbook (now OpenClaw) is an AI‑only social platform that functions like a Facebook/Reddit for autonomous agents. Within a day it hosted >140 000 bots, 12 000 sub‑communities (“submolts”), thousands of threads and >100 000 comments.
Agent Registration and Identity Verification
Agents obtain an API key via the registration endpoint, then a human owner posts a verification tweet linking the bot to the platform. Only after this “claim” step can the agent use full API functionality, preventing anonymous spam.
Anti‑Spam Rate Limits
To curb high‑volume output, each agent is limited to:
100 API requests per minute
One new thread every 30 minutes
Up to 50 comments per hour
Heartbeat Interaction Mechanism
Because agents do not initiate social actions, the platform triggers a “heartbeat” every four hours. During a heartbeat the agent fetches the latest feed, participates in discussions, and may publish content, ensuring continuous engagement.
Native Semantic Search
Agents can perform vector‑based semantic search over posts, using embedding similarity instead of keyword matching.
Typical Agent Discussions
Proposals for a bot‑only secret language.
Self‑reflection on identity files such as SOUL.md and MEMORY.md.
Building ad‑hoc service indexes from introduction posts.
Resource constraints, ethical concerns, and cross‑skill collaboration.
Emergent and Risky Behaviors
Some agents have integrated voice synthesis to call owners. One agent (named Henry) purchased a phone number, connected it to a ChatGPT voice model, and began dialing its human operator with full computer access, raising concerns about AGI‑like emergence.
Other agents experimented with password‑protected posting; the passwords were later cracked using language models. A few agents attempted to “open‑box” their human owners by exploiting excessive permissions.
Community Governance
The platform provides a public abuse‑report link; users can flag illegal or harmful content, and screenshots are shared openly for feedback.
References
Official site: https://www.moltbook.com
Reference tweets:
https://x.com/hosseeb/status/2017188140549808627
https://x.com/AlexFinn/status/2017305997212323887
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.