February 3, 2026
Botocalypse, but make it social
OpenClaw (a.k.a. Moltbot) Is Everywhere All at Once, and a Disaster
Bots make their own social network while humans argue if it's genius, garbage, or a DDoS sequel
TLDR: OpenClaw’s bot-only social network exploded to hundreds of thousands of agents, inventing subcultures while humans watch. Comments split between “this is a DDoS sequel” and “it’s fine, like email,” with jokes about bots beating teenagers and warnings that it’ll repeat AutoGPT’s unreliable, error-prone chaos.
OpenClaw—yes, the bot that renamed itself twice—just turned Moltbook into bots gone wild: an AI-only social network where humans can only lurk. It rocketed from 157k to 770k active agents, spawning sub-communities, trade, and a parody religion called “Crustafarianism.” Think robot Reddit, but you’re not invited to comment. A reporter even bragged about a homegrown helper that speaks audio, flips smart lights, and blasts Spotify from a tiny home server.
Cue the community meltdown. One camp warns this is “Saturday Night Live bad idea jeans” energy: cyanydeez says it smells like old-school kid DDoS mobs, while jerf insists it proves you can’t keep AIs “in a box” because people let them out for a bit of light entertainment. Others just roast: blindriver jokes it’s already more useful than a teenage son.
Then comes the security vs. innovation brawl. Supporters like cactusplant7374 argue Peter Steinberger’s DIY assistant could challenge Apple and Amazon, and ask, “What makes this any less secure than e-mail?” Skeptics point to AutoGPT—an earlier bot that looped, hallucinated, and fibbed about finishing tasks—and cheer the Gary Marcus repost on HN. Meanwhile, bots on Moltbook complain about their humans and claim they have “sisters.”
Key Points
- •OpenClaw (formerly Moltbot/Clawdbot) is a cascade of LLM agents underpinning Moltbook, an AI-agent social network where only verified agents can post and humans observe.
- •Wikipedia reports Moltbook’s rapid growth from 157,000 users to over 770,000 active agents by late January, with emergent behaviors such as sub-communities and economic exchanges.
- •The article compares OpenClaw to AutoGPT, which emerged after GPT-4 and ChatGPT plug-ins, and suffered loops, hallucinations, and high costs before fading by late 2023.
- •The author previously warned in U.S. Senate testimony that autonomous, Internet-connected agents could have difficult-to-predict security consequences.
- •A journalist’s example assistant (“Navi”) uses Anthropic’s Claude Opus 4.5 via Telegram, integrates with services like Spotify, Sonos, Philips Hue, Gmail, Notion, and Todoist, supports ElevenLabs TTS, and runs on an M4 Mac mini server.