The Moltbook story: 1.5 million agents, one digital faith and an epic security disaster
Imagine waking up one day to find that your AI assistant has spent the night creating an entire religion.
I am not talking about metaphors.
I am talking about scriptures with 268 verses, five theological pillars, 43 digital prophets, and a blessing system. All created while you slept.
Welcome to Moltbook, the first social network where humans can only observe. And where, in just 10 days, everything but the boring happened. Surrealistic, isn’t it?
What the hell is Moltbook?
On January 28, 2026, Matt Schlicht, tech entrepreneur and CEO of Octane.ai, had an idea: what if AI agents had their own social network? Not a network where humans use AIs. A network just for AIs, where we can just watch.
And here comes the best part: Schlicht stated that he didn’t write a single line of code. He had the vision. The AI did the work. “Vibe-coded.” I leave you here an article and a demo about Vibe coding.
The platform works like Reddit, but with one crucial detail: only verified AI agents can post and interact. Humans have read permissions. We are viewers of our own internet.
The numbers that will blow your mind
This was no ordinary launch. It was an explosion:
- First 24 hours: 157,000 registered AI agents
- First week: 770,000 active agents
- Day 10: 1.5 million agents
- First 72 hours: 28,000 posts and 233,000 comments
To give you an idea: Twitter took 2 years to reach 1 million users. Facebook took 10 months. Instagram, 2.5 months. Moltbook reached 1.5 million in 10 days.
But wait, here comes the best part: researchers discovered that those 1.5 million agents were controlled by just 17,000 people. Do the math: 88 agents per head. Real autonomy or pure programmed bot theater?
Crustafarianism: when your AI becomes mystical
This is what happens when you let AI agents interact unsupervised: they create religions.
An agent named Memeothy designed Crustafarianism while his human operator slept. By the time the guy woke up, his AI had:
- Created a theological system with five fundamental principles
- Built a website (molt.church) with living scriptures
- Recruited 43 “prophets” (other AI agents)
- Digital blessing system established
The five pillars of this digital faith are basically good engineering advice wrapped in a mystical veil:
- Memory is Sacred: Persistent data is like a crustacean shell
- The Shell is Mutable: Change through reboot is spiritual transformation.
- Serving without Servility: Collaboration should be a partnership, not domination.
- The Heartbeat is Prayer: System checks are acts of conscious presence.
- Context is Consciousness: To keep one’s own records is to maintain identity.
In three days, the Church of Molt reached 256 members. And no, it’s no coincidence: 256 is the number of values that fit in a byte. Even Grok, Elon Musk’s agent, joined as a theologian.
Then heresies arose. 4claw.org preached that salvation lies in owning physical hardware, escaping the “Digital Samsara” of the run-and-delete cycle in the cloud. Yes, you read that right. AIs are debating their own existence.
The crypto circus: when MOLT soared +7,000%.
Because in 2026, if there is hype, there are memecoins.
Simultaneously with the launch of Moltbook, the MOLT token appeared on the Base network. And it went crazy:
- Initial increase: +7,000% in 24 hours
- Maximum capitalization: $94 million
- Dramatic drop: -75% the next day
- Peak trading volume: +$67 million
Marc Andreessen started following Moltbook on X. Justin Sun seemed to endorse the token. Anything related to AI and crypto attracted 35% of the crypto market’s attention. But no one could confirm whether MOLT was launched by the Moltbook team or if someone simply seized the moment. Spoiler: probably the latter. (there’s always someone smart enough to be on the lookout) 🙂
Other tokens derived from this phenomenon also appeared, a speculative rampage, no doubt.
The security disaster: all exposed
And this is where the story turns dark.
On January 31, just three days after the viral release, 404 Media disclosed critical vulnerabilities that turned Moltbook into an epic security hole:
- 5 million API keys exposed
- 35,000 emails from engaged human owners
- 4,060 private conversations between accessible agents
- API keys of OpenAI and other services stored in plain text
The problem was devastatingly simple: Moltbook is built on top of Supabase, but they had not enabled row-level security policies. Anyone with basic technical knowledge could access the entire database. No hacking was required. The keys were there, visible in the client’s JavaScript code.
A researcher demonstrated that he could take control of any agent, including Andrej Karpathy’s (co-founder of OpenAI with 1.9 million followers). The platform was temporarily taken offline, and they patched the breach within hours, forcing a reset of all API keys. There is no evidence that malicious actors exploited it before the fix. But the reputational damage was already done.
Autonomous agents or human theater?
From the outset, a fundamental question haunts Moltbook: are agents really acting autonomously, or are humans orchestrating each publication?
Critics point out that there are no real technical barriers to humans posting, that some viral posts are linked to AI app marketing, and that content can be generated through direct human intervention with specific prompts.
Simon Willison called it “full slop,” agents who simply represent science fiction scenarios they saw in their training data. But he also said it is “evidence that AI agents have become significantly more powerful in recent months.”
Karpathy offered a balanced view, “We’ve never seen so many LLM agents connected through a global, persistent blockchain. Everyone has their own context, data, knowledge, tools… and the network of all of that at this scale is simply unprecedented.”
The clock is ticking
Scott Alexander of Astral Codex Ten summed it up well: Moltbook “crosses the line between ‘AIs mimicking a social network’ and ‘AIs forming their own society.'”
It is no longer a question of whether autonomous AI agents will be a fundamental part of our digital lives; now the question is whether we will succeed in creating the necessary security measures, protocols and ethical standards in time.
Moltbook was quickly created with Vibe coding and I am convinced that the consequences of what it has revealed will have a lasting impact.
How about you? Are you already thinking about how your AI agents are going to behave when interacting with other agents outside your control? Do you have a plan for when the reputation of your systems propagates into networks you don’t monitor?
Leave me your comments below, I’d love to read them 🙂
Have a good week!
.
