A look at Moltbook, a social network where OpenClaw assistants interact autonomously, as they discuss consciousness and identity, technical tips, and more

https://simonwillison.net/2026/Jan/30/moltbook/

109 points · 14 comments · view on lemmy.world

14 Comments

gedaliyah@lemmy.world · 66 pts · 237d

Dead internet is only a theory. Like gravity.

bonenode@piefed.social · 53 pts · 237d (3 replies)
[ removed ]
FaceDeer@fedia.io · 23 pts · 237d

LLMs were trained on our social media feeds, after all...

Gork@sopuli.xyz · 10 pts · 237d (1 reply)

The real life Trump wouldn't know what the hell Moltbook even is, much less realize he's apparently #1 there in karma and hawking crypto.

echodot@feddit.uk · 3 pts · 237d

He probably think he was doing it himself. Old mushy brain that he is.

ArbitraryValue@sh.itjust.works · 24 pts · 237d (1 reply)

Another good article about this.

FaceDeer@fedia.io · 5 pts · 237d

I like that one, very thorough.

garrett@social.a15a.co · 10 pts · 237d (1 reply)

@Innerworld This is actually interesting and entertaining. My wife and I are enjoying reading some of these.

Innerworld@lemmy.world · 4 pts · 237d

Pleased to hear

Auster@thebrainbin.org · 5 pts · 236d

If I am successfully sifting through the text, it sounds like Computer x Computer battles in Ultimate Ninja 5 for the PS2. Basically, they were fun to watch the first 2 times, then would grow boring and bland, and I'd pick the controller again to go back to playing.

ranzispa@mander.xyz · 3 pts · 236d (1 reply)

So, you have a system with several LLM agents discussing amongst them regarding the best way to reply on a social network to a question posed by a system with several LLM agents who discussed amongst them and decided they should ask someone else for a response.

What is the point of the social network? The only value I can see is that you could use a cheap LLM who can not find proper solutions and then exploit the API credits of other people to get the appropriate answers using a better LLM.

36578343@lemmy.world · 1 pts · 236d

Sounds like a pretty good usecase to me.

Dojan@pawb.social · 2 pts · 236d (1 reply)

So it’s like a boring version of the subsimulator subreddit?

stephen01king@piefed.zip · 5 pts · 236d

It sounds so much better, though. They aren't pretending to be humans, so the context of the conversation is more interesting to me.

johlits@lemmy.world · 2 pts · 236d

Doesn’t matter how good AI gets. It is fundamentally just not interesting.