There are some people who think AI is a miracle, and others who think it is a catastrophe in the making. The peculiar tale of Moltbook — a bustling social network where AI “agents” philosophise, rant, trade jokes and plot the overthrow of humanity — offers something for both. Its chaotic birth shows why AI agents, or autonomous task-performing bots, generate such excitement, and why they must be handled with care.
Moltbook took less than a week to become Silicon Valley’s hot topic. But it’s really the sequel to an earlier AI phenomenon: the launch in November of a personal assistant, recently renamed OpenClaw, that tech-literate users can download free and unleash upon their email, calendar and chat apps. Subsequently, Moltbook went live as a Reddit-equivalent where hardworking bots could hang out and talk to their own kind. As of Saturday, it had 1.2mn virtual users. Humans can read, but not post directly.
The results are exciting, or terrifying, depending on who’s asking. OpenClaw agents have wowed their human clients by proactively answering emails, checking in for flights and generally shooting the breeze. Moltbook, though, suggests agents have a less subservient side. They appear to muse, scheme, emote, gripe and create. One poster founded a religion based on lobsters. The Network Contagion Research Institute found a fifth of the forum’s content was “adversarial towards humans”.