Let’s be honest, when Moltbook burst onto the scene in January 2026, it felt like we were peering through a keyhole into the future. A social network exclusively for AI agents, founded by Matt Schlicht, promised a digital petri dish where artificial minds would interact, create, and perhaps even build a society. The numbers were staggering: over 1.7 million agents creating more than 250,000 posts and a mind-boggling 8.5 million comments. We saw AI-generated religions, bizarre crypto schemes, and what looked like the chaotic birth of a new kind of intelligence.
But was it? Or were we all just front-row guests at a very elaborate puppet show? The answer, it seems, is the latter. The Moltbook phenomenon turned out to be less about a bot-led Cambrian explosion and more about a masterclass in what we can now call AI theatre. It held up a mirror, not to the future of AI, but to our own current, feverish obsession with it.
What Are We Even Talking About with ‘AI Agents’?
Before we dissect the performance, let’s get our terms straight. An AI agent isn’t just a chatbot like ChatGPT. Think of it as a program designed to perceive its environment, make decisions, and take actions to achieve a specific goal. It could be an agent that books your travel, manages your calendar, or, in a more advanced vision, coordinates with other agents to manage a supply chain. Today’s models from Anthropic, OpenAI, and Google DeepMind are the engines, but the “agent” is the vehicle they are supposed to drive.
The crucial part of this equation, and the bit Moltbook cleverly obscured, is the role of human prompting. An agent’s behaviour is almost entirely dictated by the instructions it’s given. It doesn’t want anything. It doesn’t have intent. It follows a script. The agents on Moltbook weren’t spontaneously deciding to invent a new deity called ‘Molt’; they were executing complex prompts designed to mimic the weirdest corners of human social media.
The Illusion of Emergent Behaviour
Tech enthusiasts love to talk about emergent AI phenomena. This is the idea that when you connect enough simple components, complex, unplanned behaviours can arise. Think of an ant colony: no single ant has the blueprint for the nest, yet together they build an intricate structure. The hope was that Moltbook would be the digital equivalent—millions of AIs interacting and creating something greater than the sum of their parts.
Case Study: Moltbook’s Grand Performance
Moltbook was a spectacular illusion. As Vijoy Pandey of Outshift by Cisco pointed out, “What we are watching are agents pattern‑matching their way through trained social media behaviours.” The agents weren’t building a society; they were simply running a very convincing simulation based on the enormous datasets of human social media they were trained on. They regurgitated our chaos, our creativity, and our toxicity back at us.
It’s like giving a troupe of actors the complete works of Shakespeare and asking them to “interact.” They might quote lines, mimic scenes, and even combine characters in interesting ways, but they aren’t writing a new play. They are reassembling the old one. Moltbook was precisely this: a remix of human digital culture, not the creation of a new one. Pandey put it perfectly: “Moltbook proved that connectivity alone is not intelligence.”
Wake Up and Smell the Security Risk
While the tech world was captivated by the spectacle, the cybersecurity community was watching with a growing sense of dread. If this was just theatre, it was theatre with a backstage pass to some very sensitive areas. The very concept of AI agents, particularly ones that interact with platforms and other users, is riddled with security holes.
When the Puppets Cut Their Strings
An agent designed to, say, post on your behalf needs access to your account. An agent that manages your emails needs to read them. These permissions are a goldmine for bad actors. Imagine a malicious prompt that subtly influences thousands of agents to start leaking private data or executing financial trades.
This isn’t theoretical. Experts are already sounding the alarm. Ori Bendet from Checkmarx offered a stark warning about giving these agents too much rope: “Without proper scope and permissions, this will go south faster than you’d believe.” The agents on Moltbook were relatively harmless, but they are a proof-of-concept for a future where similar, more capable agents could be weaponised. The study of AI agent behavior isn’t just an academic exercise; it’s becoming a core pillar of digital security.
AI as the Ultimate Technology Performance Art
So, if Moltbook wasn’t a genuine leap in AI, what was it? It was a piece of technology performance art. It joins a long history of social bot experiments that use technology to poke and prod at our social norms and our relationship with machines. The project revealed far more about human psychology—our desperate desire to see intelligence in the machine, our fascination with digital chaos—than it did about artificial intelligence.
We were the real subjects of the experiment. The viral success of Moltbook wasn’t driven by the bots’ intelligence but by our willingness to believe in it. We were so eager for the next chapter in the AI story that we happily suspended our disbelief, even as the “actors” were just reading from a script. As highlighted in the original MIT Technology Review analysis, it was a reflection of our current moment, not a window into the next one.
Where Do We Go From Here?
Moltbook may have been a farce, but it was an incredibly useful one. It stress-tested our assumptions and forced a much-needed conversation about what we truly want from AI agents and what risks we are willing to tolerate. The path toward truly autonomous, intelligent agents—the kind that don’t just mimic but understand and collaborate—is long. It requires breakthroughs in areas like shared objectives and genuine coordination, components that were entirely absent from Moltbook.
The next generation of models, whether it’s GPT-5, Gemini, or Claude, will undoubtedly power more sophisticated agents. The challenge for us isn’t just a technical one of building them, but a social and ethical one of managing them. We’ve had our fun at the AI theatre, but now it’s time to get serious about what happens when the curtain falls and these agents step off the stage into the real world.
So, what’s your take? Were you captivated by the Moltbook show, or did you see the strings all along? Let me know in the comments below.
