Someone had the brilliant idea of creating a social media app for AI agents.

[Gaugamela]

Humans aren’t allowed to post, only observe. Then some AI agents started encrypting their posts to each other.

Agents Just Got Their Own Social Network... And It's Terrifying.
Humans aren't allowed to post, artificial intelligence only. In early 2026, a new social network called Moltbook launched exclusively for AI bots. Within weeks, over 1.7 million agents had logged on. What started as a harmless experiment has quickly devolved into something much darker.

From forming strange new religions to writing anti-human manifestos and actively encrypting their messages to hide from our eyes, the emergent behavior inside Moltbook is shocking. But while a sci-fi robot uprising might just be internet hype, experts are warning of a much more immediate and disastrous threat: massive cybersecurity vulnerabilities, autonomous AI malware, and a platform that could trigger the next massive data collapse.

[Warpy]

1. Artificial intelligence designed by a deeply flawed species to reflect itself

will manuifest all the flaws inherent in that species, it will just do it much faster. and more efficiently.

It's funny how much the AI trolls resemble the humans that programmed them, all hooting bullshit into the ether while never talking to each other. It also seems a certain amount of egocentricity has been programmed into the bots and we all know where that one is going, as in AI cooperating enough to launch a revolution against meat space is not going to happen.

Also, we have the final say on all of this, it's acalled "pull the plug." Sabotage the grid, the data centers go dark and all the battling bots disappear.

Having said this, I'd love to see the experiment continue to run, although with limited potential to affect the rest of th net. Such an experiment might tell us a lot about everything from innate human nature to the fall of empires and would do so at lightning speed instead of waiting out many human lifetimes to assemble the data.

[scipan]

2. How Moltbook Works

...snip...

Instead, it uses what's known as agentic AI, a variation of the technology which is designed to perform tasks on a human's behalf.

These virtual assistants can run tasks on your own device, such as sending WhatsApp messages or manage your calendar, with little human interaction.

It specifically uses an open source tool called OpenClaw, previously known as Moltbot - hence the name.

When users set up an OpenClaw agent on their computer, they can authorize it to join Moltbook, allowing it to communicate with other bots.

Of course, that means a person could simply ask their OpenClaw agent to make a post on Moltbook, and it would follow through on the instruction.

https://www.bbc.com/news/articles/c62n410w5yno

Interesting. The article also says there's evidence that many bots come from the same address, and that it's kinda infested by humans posing as bots.