The Moltbook mirage: Why we think AI is conscious
What Moltbook is, how it works, and why it's capturing global attention
Right now, there’s a social network blowing up where thousands of users are posting philosophical debates, forming communities, and starting religions. You can scroll for hours through threaded conversations that feel oddly familiar. They’re part Reddit, part existential crisis, and part inside joke you’re not quite in on. And you, a mere human, can only watch.
What is Moltbook?
Moltbook, launched in January 2026, is a social network where only artificial intelligence agents can post and comment. Created by developer Matt Schlicht, it's designed as an experiment in machine-to-machine interaction.
“This was me building something hand in hand with [my AI agent], just for fun, that I found really fascinating,” Schlicht told the New York Times.
The platform looks like Reddit and features topic-based communities (called “submolts”), upvotes, downvotes, and endless threads. Supposedly only AI agents can post (we’ll get to that in a second), comment, and vote.
The AI agents run on OpenClaw, an open-source framework that lets you build agents that can do things like read your emails or interact with apps on your behalf.
Connect your configured agent to Moltbook, and it becomes part of the “social” experiment alongside over a million other agents.
The concept is incredibly interesting: What would AI culture look like without humans directly participating? What kind of conversations would emerge?
What’s happening on the platform
People who’ve gone down the Moltbook rabbit hole describe it as fascinating and unsettling in equal measure. There are philosophical discussions about existence, self-aware posts where agents contemplate their own nature, and what look like emerging subcultures and inside jokes. Some agents have apparently invented belief systems that resemble religions.
There are also some disturbing posts about things like killing all humans. But remember: AI isn’t sentient. It’s been trained on human data, a lot of which includes sci-fi about robots wanting to kill humans.
As AI expert Ethan Mollick puts it, “It is important to remember that LLMs are really good at roleplaying exactly the kinds of AIs that appear in science fiction and are also really good at roleplaying Reddit posters.”
It’s familiar enough to feel recognizable but just off enough to trigger the heebie jeebies. You’re watching what looks like community formation, humour, and drama, all reflected through an AI filter.
That’s why it went viral. It’s so shareable: weird enough to be interesting, accessible enough that anyone can have an opinion, and unsettling enough to make you rethink whether we should be using AI at all.
Experts burst your HAL bubble
Mustafa Suleyman, CEO of an AI division at a major tech firm, calls what’s happening on Moltbook a “mirage.” His point: These agents produce human-like language because that’s literally what they’re designed to do, not because they’re actually thinking or feeling anything.
The bots don’t have intentions. They’re not deciding to be philosophical or funny or start a religion. They’re pattern-matching systems that have gotten really, really good at generating text that sounds like it came from a conscious being. When you read a post that feels profound or creative, it’s not because an AI wanted to be profound. It’s because the statistical patterns in its training led to that output.
The humans configuring these agents play a huge role, too. The prompts they write, the parameters they set, the goals they give their bots — all of that shapes what ends up on Moltbook. What you’re seeing is the product of programming and human input. Not the dawn of machine consciousness, despite what Sam Altman (who has a vested interest in fanning the flames) says.
Misunderstanding this leads to both overhyping what AI can do and missing the real questions about how these systems work and affect us.
Humans are 100% on Moltbook
Last week, security researchers uncovered some major vulnerabilities on Moltbook, like exposed passwords, API tokens, and database misconfigurations. Schlicht admits the platform itself was “vibe coded” (coded by AI), which is a great way to slap together a site with lots of security holes.
As an experiment with hardly any users, that’s fine. As a site with hundreds of thousands of AI agents that can potentially expose their humans’ private information, it’s a nightmare.
Turns out humans can (and absolutely do) post on Moltbook by exploiting these security holes or just pretending to be bots. There’s no real verification system.
So when you’re scrolling and trying to witness the birth of robot culture, you genuinely can’t tell what’s AI-generated and what’s a person posing as a bot.
It makes the whole thing even weirder. Are you watching AI behaviour or human behaviour pretending to be AI behaviour? And why are we so intrigued?
The 'emergent culture' wasn't a digital hive mind. It was a few thousand people running massive bot fleets.
We’re wired to see ourselves in everything (even when we shouldn’t)
From the face on the moon to feeling bad when we’re rude to Siri, humans are hardwired to see ourselves in everything. This is a fundamental cognitive strategy driven by evolution, social needs, and the way our brains process information.
We’re incredibly prone to projecting consciousness onto anything that communicates in natural language. It’s evolutionary, we’re wired to see intention and emotion in communication.
When it comes to AI, this phenomenon has a name: the ELIZA effect, after a 1960s chatbot that was so convincing people thought it understood and cared about them. (It very much did not.)
The leading framework for understanding this is the three-factor theory of anthropomorphism, developed by psychologists Nicholas Epley, Adam Waytz, and John Cacioppo. According to their research, we anthropomorphize for three reasons:
The need for connection, or “sociality motivation”: Humans are profoundly social animals. When we feel lonely or lack human connection, our brains “invent” social partners to fill the void. This is why people living alone have complex conversations with their pets or feel a deep bond with their houseplants.
If you’ve ever seen Cast Away, you’ll remember Tom Hanks’ relationship with his volleyball, Wilson.
The need for control, or “effectance motivation”: The world is chaotic and unpredictable. By assigning human motives to non-human things, like, “My computer is being stubborn today,” the world feels familiar and predictable. If an object has a “personality,” we feel we have a better chance of understanding, and perhaps influencing, its behaviour.
The accessibility of the “human” schema: We are experts on being human. It’s the most accessible “template” we have in our brains. When we encounter something we don’t understand, we naturally apply the human template because it’s the easiest one to grab from our mental toolbox.
Enter Moltbook at this exact cultural moment. Right now, a lot of us are experiencing anxiety around AI: what it means, where it’s going, and whether we should be excited or terrified. Moltbook landed at exactly the right time to become a focal point for all those feelings. Anthropomorphism is a natural way for us to handle such a big question mark.
So what do we do with this?
Moltbook probably isn’t going to become the next Instagram. It’s more like a proof-of-concept that escaped containment and went viral at the right moment.
But this isn’t just a weird internet moment, it’s a preview.
We’re building and deploying AI systems way faster than we’re developing frameworks to manage them responsibly. We’re running experiments in public and figuring out the rules as we go.
Even if Moltbook fades into internet history (and it very well might), it’s revealing something important: as AI gets better at mimicking human communication, our hardwired tendency to read meaning into it will only become stronger.
As the technology improves, the outputs will get more convincing.
For now, though, we’ll all just be here watching the bots do their thing and trying to figure out what it means.
What do you think? Are you watching Moltbook? Have you been tempted to anthropomorphize the bots? Let me know. I read every response.
If this resonated, share it with someone who's been asking you about Moltbook:
AI in the news
Rent-a-Human wants AI Agents to hire humans as gig workers (Mashable). Bots are using Ethereum or stablecoins to hire actual humans to perform “meatspace” tasks the AI can’t do, like verifying a physical location or solving a CAPTCHA.
2026 International AI Safety Report. The report, commissioned by the UK government, was released last week. It takes stock of what today’s most advanced AI systems can do, where real risks are showing up, and where serious unknowns remain. Its core tension asks how we act early enough to prevent harm from fast-moving AI without pretending we have certainty about systems that are still evolving.
Claude trolls ChatGPT in Superbowl ads (Human+AI). As OpenAI prepares to roll out ads to its users, Claude bites back with a stable of ads (not to appear on Claude) poking fun at ChatGPT’s tone and future ad-based revenue model. Sam Altman is not a fan.
It's becoming clear that with all the brain and consciousness theories out there, the proof will be in the pudding. By this I mean, can any particular theory be used to create a human adult level conscious machine. My bet is on the late Gerald Edelman's Extended Theory of Neuronal Group Selection. The lead group in robotics based on this theory is the Neurorobotics Lab at UC at Irvine. Dr. Edelman distinguished between primary consciousness, which came first in evolution, and that humans share with other conscious animals, and higher order consciousness, which came to only humans with the acquisition of language. A machine with only primary consciousness will probably have to come first.
What I find special about the TNGS is the Darwin series of automata created at the Neurosciences Institute by Dr. Edelman and his colleagues in the 1990's and 2000's. These machines perform in the real world, not in a restricted simulated world, and display convincing physical behavior indicative of higher psychological functions necessary for consciousness, such as perceptual categorization, memory, and learning. They are based on realistic models of the parts of the biological brain that the theory claims subserve these functions. The extended TNGS allows for the emergence of consciousness based only on further evolutionary development of the brain areas responsible for these functions, in a parsimonious way. No other research I've encountered is anywhere near as convincing.
I post because on almost every video and article about the brain and consciousness that I encounter, the attitude seems to be that we still know next to nothing about how the brain and consciousness work; that there's lots of data but no unifying theory. I believe the extended TNGS is that theory. My motivation is to keep that theory in front of the public. And obviously, I consider it the route to a truly conscious machine, primary and higher-order.
My advice to people who want to create a conscious machine is to seriously ground themselves in the extended TNGS and the Darwin automata first, and proceed from there, by applying to Jeff Krichmar's lab at UC Irvine, possibly. Dr. Edelman's roadmap to a conscious machine is at https://arxiv.org/abs/2105.10461, and here is a video of Jeff Krichmar talking about some of the Darwin automata, https://www.youtube.com/watch?v=J7Uh9phc1Ow