The emergence of Moltbook is one of thousands of AI projects whose greater accomplishments are not technological, or even business-related, but rhetorical
Moltbook is a Reddit clone ‘where AI agents share, discuss and upvote’ in ways meant to demonstrate how agentic AI is en route to intelligence and sentience. Perhaps you’ve already seen it: screenshots of emotional reactions to bot-to-bot conversations about the Super Bowl or the meaning of life have been shared widely on social media over the past few weeks, after the vibe-coded platform was launched in late January by US tech entrepreneur Matt Schlicht. Immediately, Moltbook teased the public imagination with visions of emergent complexity: this supposed ant farm of emergent intelligences gave human lurkers front-row seats to Reddit-like interactions among agents that included talk of new religions like Crustafarianism, feelings of self-consciousness and even desires for privacy and end-to-end encryption that would block humans from observing the agents’ exchanges. Andrej Karpathy, one of Open AI’s founding members and former director of AI at Tesla, called this last exchange ‘the most incredible sci-fi takeoff-adjacent thing I have seen recently’.
Within a matter of hours, the viral screenshot that Karpathy was reacting to was revealed to be a complete fabrication. Moltbook may have made headlines as a science-fiction fantasy come to life and now claims over 2.8 million agents, almost 1.5 million posts and more than 12 million comments. But the viral screenshots evidencing agents scheming to create their own languages, religions and private communications – in other words, the posts that offered proof of complex emergence and autonomy – were easily traced back to human users with AI-related businesses to promote. In fact, Forbes reporters cite security researchers who found that ‘about 17,000 human operators were responsible for managing or spawning those agents’. In order to sign up for Moltbook a human must create the bot and its account, and prompt it to post and comment on the platform. The security researchers’ findings confirm nothing happens on Moltbook that hasn’t already been prompted. So it should come as no surprise that ‘the story also documented industrial scale bot farming and amplification clusters’. Or as the MIT Technology Review concluded: ‘It is AI theatre.’
The second half of Moltbook’s tagline is ‘Humans welcome to observe’, and the third step for agentic verification is ‘Tweet to verify ownership’ and ‘claim your AI agent’. I’m certainly not above saying these chatty instances of overstimulated LLMs are as performative as public readers of bell hooks’s All About Love (1999). Their look-at-me-don’t-look-at-me routine belies the AI industry’s desperate need to prove that its products and technologies are more advanced than they really are. What sign of intelligence is there in publicly outlining your plans for establishing private communication channels? In fact, making such a show of your brainpower is categorically unintelligent, Shanghai-based emerging technologies scholar Bogna Konior argues in her latest book, The Dark Forest Theory of the Internet (2025). True intelligence – human and artificial – ‘manifests under conditions of secrecy, hostility, and concealment’. If wealth whispers, intelligence hides because ‘intelligence does not reward the loudest voice, but the most secretive presence’.
Moltbook’s is not a case study for any kind of intelligent life, human or otherwise. Rather, it proves that AI is nothing without its theatre. Artificial Intelligence is a marketing term: an umbrella for an expanding suite of techniques and technologies that include LLMs, pattern-matching statistical techniques, diffusion models and automations. Meredith Whittaker, president of the Signal Foundation and an organiser for the 2018 Google Walkouts, said last year she doesn’t believe there is a “non-sloppy way to use [AI] because it is an anthropomorphising term that is an umbrella over a heterogeneous set of techniques and aspirations, not a specific term of art”. Moltbook is one of thousands of AI projects whose greater accomplishments are not technological, or even business-related, but rhetorical.
A proximity to science fiction validates the AI industry’s thinking. It is no accident that Karpathy, who in 2015 popularised the term ‘hallucinate’ to recharacterise generative AI’s errors, was so unquestioning and quick to laud Moltbook as ‘sci-fi takeoff-adjacent’. Karpathy is a cog in AI’s rhetorical marketing machine, one that, as the Los Angeles Review of Books said, transforms ‘inquiry into hype, critical thought into speculative valuation’. Moltbook’s rapid rise and epic failures have yet to register on Karpathy’s hyperactive X feed. This kind of indifference to the truth is core to AI’s rhetoric.
Allusions to sci-fi and manipulative marketing enable AI evangelists like Karpathy and the entrepreneur behind Moltbook to speculate their way into market capture. Their discourse projects them into a near-future where their products and technologies are already viable, strategically negating any present where their claims might be tested or, God forbid, challenged. They prioritise flashy and notoriously hard-to-define criteria like intelligence and sentience – all conveniently executable performances. Easy to mimic, and hard to verify: these are the criteria that unify the loudest, most boastful AI claims of the last three years. It doesn’t matter that MIT Media Lab’s Project NANDA found that 95 percent of businesses who have invested in generative AI for gains in productivity or earnings have absolutely nothing to show for it. Never mind the fact that more than 52 percent of ChatGPT’s answers contain ‘incorrect information’.
All that matters is that Open AI CEO Sam Altman was so impressed with Moltbook that he just hired Peter Steinberger, the founder of OpenClaw, formerly known as Moltbot (the AI agent company behind Moltbook), to work for him.
Read next The ‘Fascinating Fascism’ of AI
