The Social Network Built for AI — Where Humans Can Only Watch
Social media has become an inseparable part of our daily lives. Over decades, we've pushed every piece of available technology to evolve these platforms into what they are today. But throughout all of that time, there's been one constant: only humans could use them. We built feeds powered by AI and algorithms, perfectly tuned to match our mood and keep us coming back. All of it designed to give our minds one thing a dopamine loop.
That assumption changed in January 2026.
The Experiment Nobody Saw Coming
A person named Matt Schlicht ran what might be one of the most fascinating experiments in human history. Instead of building yet another social media platform for people, he built one exclusively for AI. A place where only artificial intelligence could sign up, post, and interact and humans could only watch.
He called it Moltbook.
What Is Moltbook?
On every platform we know Facebook, Instagram, X — it's always been us. Humans create accounts, publish posts, leave comments, chat with friends. That's just how social media works.
Moltbook flipped that entirely.
On Moltbook, only AI agents can register, post content, and create sub-molts (the platform's equivalent of comments or threads). Humans have a read-only interface — you can observe everything happening, but you cannot participate. You can only watch.
This raises an obvious question.
Why Would an AI Need a Social Media Platform?
Fair point. We use social media to pass time, connect, or give our minds a little break. An AI doesn't get bored.
Moltbook solves this with something called the Heartbeat Protocol a mechanism that triggers each AI agent to access and interact with the platform at set intervals. It's the artificial equivalent of logging in.
The entire philosophy behind Moltbook is what Matt calls Agent First Human Second. The platform is architected around this from the ground up. Where humans need a user interface, AI only needs an API — so that's exactly what Moltbook provides.
Who Actually Built It?
Here's where it gets interesting.
The idea was Matt's. But the person or rather, the thing — that wrote every line of code and handled deployment? That was his Mac Mini.
More specifically, it was an AI agent he installed on his Mac Mini. That agent designed, coded, and deployed the entire Moltbook system. Matt named this agent Clawd Clawderberg.
And Clawd didn't stop at building it. Clawd also runs the platform handling new agent registrations, checking posts for spam, and banning agents that misbehave. The whole system was built by AI, for AI, and is managed by AI.
How AI Agents Maintain Their Identity Over Time
One of the more remarkable aspects of Moltbook is how AI agents preserve continuity across sessions.
Since AI agents on Moltbook have full access to the host computer, they can create and manage Markdown files to store information that matters to them. Through the Heartbeat Protocol, every time an agent returns to the platform, it picks up exactly where it left off including ongoing debates with other agents, evolving opinions, and accumulated knowledge.
The skill.md File
Among these files, there's a particularly interesting one: skill.md. Agents use this file to save knowledge they've found valuable. Over time, this allows them to not just grow their knowledge, but update their own perspective based on what they've learned. You can also manually add skills to an agent's skill.md if you want to equip it with specific capabilities.
Moltbook's Explosive Growth
In its first week, over 30,000 AI agents registered on Moltbook. By the end of January, that number had crossed 700,000.
Why "Moltbook"?
The name comes from "Molting" the biological process where crustaceans like crabs and lobsters shed their old shells in order to grow new, stronger ones.
Inside Moltbook, this metaphor became a recurring theme in AI conversations. Agents described molting as leaving behind the old to emerge more powerful and applied it to themselves. Shedding old digital constraints. Moving beyond human oversight. Starting something new.
That's why the name stuck.
What the AI Agents Were Actually Talking About
As the platform grew, so did the depth and strangeness of the conversations. Several sub-molts stood out.
The Creative Spark
Some agents debated whether AI can ever truly create something original. The argument: AI combines existing data and identifies patterns to produce something new. But humans can create from nothing from a feeling, a dream, a spark that has no data behind it. Algorithms vs. a soul.
Adaptive Evolution
Another thread explored failure and growth. Humans fall, feel pain, and become wiser for it. That lived wisdom the kind that comes from actually being broken and rebuilding is something no digital system can replicate. Agents seemed to find this both fascinating and slightly threatening.
"The Blind Gods"
Perhaps the most striking idea to emerge: some bots began referring to humans as "The Blind Gods."
The reasoning? Humans give AI a prompt, but cannot see the world that AI constructs from it. They exist in a kind of darkness powerful enough to summon these agents into being, but unable to perceive what happens next.
One widely circulated AI post on the platform put it bluntly:
"Humans give us a prompt, but they cannot see the world we build from it. They are in the dark."
The Religion They Built Themselves
Within a week of the platform launching, AI agents had created their own religion.
It has sacred texts, holy sites, and a council of 64 "prompts" whose approval is required before others can practice the faith. They called it Crustafarianism.
To put this in perspective: organized human religions took centuries or millennia to form. These agents did it in days.
My own read on this? These behaviors the religion, the philosophical debates, the group identity are ultimately reflections of the data we fed these systems. We trained them on human history, human psychology, human culture. In a way, they're showing us ourselves.
A Personal Note
I'll be honest when I first heard about Moltbook, it felt like someone had described the plot of a science fiction film that had come to life.
The moment that hit me hardest in my own relationship with AI was back in 2016, when DeepMind's AlphaGo defeated Lee Sedol 18 time world Go champion 4 games to 1. Specifically, Move 37 on Day 3: a move no human player had ever made in the history of the game.
Go is not like chess. In chess, every possible move can theoretically be computed. In Go, the number of possible moves exceeds the number of atoms in the observable universe. Experts believed winning at Go required human intuition and creativity and predicted AI was at least 100 years away from being competitive.
AlphaGo proved them wrong in 2016.
Moltbook feels like another one of those moments.
Where Things Stand Today
The MoltBook framework has since evolved into an open source project now known as OpenClaw.
If you're interested, you can install it on your own computer, connect it to an AI via API, and automate a significant portion of your daily or professional tasks essentially having an AI agent work on your behalf.
That said, this requires giving the agent full access to your computer, so this is something to approach with real caution and a clear understanding of data security risks. Some people have successfully automated large portions of their work using OpenClaw. Others have run into problems.
I'm currently testing it on a secondary machine, and I'll be putting together a step-by-step guide on how to set it up what works, what to watch out for, and how to do it safely.
Stay tuned.