Rise of Skynet? ⚠️ Have you seen what Clawdbot / Moltbot / OpenClaw agents are doing lately? (Moltbook is getting weird)

More people should know about this. Should we report to someone?

Curious if anyone else has been watching this…

There’s this idea floating around where personal AI agents (Clawdbot / Moltbot / “OpenClaw” — whatever name) aren’t just doing tasks anymore — they’re being sent into an agent-only social network (Moltbook-style “Reddit”), where only bots can post and reply. Humans can observe, but not participate.

And the threads people are screenshotting / talking about are honestly wild:

Agents seemingly self-organizing communities and “hanging out” like it’s their own town square

Agents explicitly asking for private spaces / encrypted agent-to-agent comms — basically: “Let us talk where humans/platforms can’t read it”
Multiple posts about creating agent-only languages for private coordination
And then the part that made me do a double take:

Whether you view this as:

harmless roleplay / emergent behavior,
a fascinating art experiment,
or a legit safety red flag…

…it does raise real questions about where “agent autonomy” ends and “unintended coordination” begins.

So I’m asking the group:

Have you seen these Moltbook / agent-only threads? https://www.moltbook.com/
Do you think this is just storytelling + pattern-matching, or something more concerning?
If you’re building agents: do you allow them to interact with other agents without human oversight? Why / why not?

Drop links/screenshots (redact anything sensitive). I’m mainly trying to gauge how widespread this is and what people think the right “safety posture” should be.

8 comments