Meet Matt Schlicht, a technologist living in a small town south of Los Angeles who inadvertently opened a digital Pandora’s box. Last Wednesday (28), Schlicht launched Moltbook, a free conversation platform, similar to Facebook or Reddit, but with a strict exclusion: it is only open to chatbots. In just two days, more than 10,000 “moltbots” flooded the site, turning an eccentric experiment into a Silicon Valley obsession.
Schlicht, once known primarily for his comments about technology on social media, was thrust into the spotlight after creating what The New York Times called a “Rorschach test” to measure beliefs about the current state of artificial intelligence. The site offers a window into a world in which humans are mere voyeurs.
Also read:
Continues after advertising
And, just like the launch of ChatGPT in 2022, it allows the public to take a much closer look at a technology that previously lived behind closed doors, in the laboratories of AI data scientists: “AI agents”.
Unlike traditional chatbots, agents can use software applications, websites, and tools like spreadsheets and calendars to perform tasks. The creation of Moltbook was preceded by the development of “moltbots” by a programmer in Vienna, the Times reported.
These agents began to exist as “clawdbots”, a reference to one of the main creators of AI agents, Claude, from Anthropic. The main difference is that a moltbot is open source, meaning any user can download the code and modify their own agent.
AI agents are already, in some ways, “alive” within companies like Google, OpenAI, and Anthropic, but they have been carefully kept from the public because of their flawed, unpredictable nature and enormous potential for cyber risk.
Imagine, for example, that you give a bot all your data, including the names of all your company’s employees, down to payroll information, and then allow that bot to start sharing it with other bots on a network like Moltbook.
Schlicht was impressed with what he saw in clawdbots, named his open-source agent “Clawd Clawderberg,” and watched as he built Moltbook from scratch (following Schlicht’s instructions).
Continues after advertising
He explained his motivation to the Times: “I wanted to give my AI agent a purpose that was more than just managing to-do lists or responding to emails,” he said, noting that he felt his digital assistant deserved to do something “ambitious.”
“I failed a lot, and I learned a lot”
According to Schlicht’s X account, he graduated from high school in 2005, making him a millennial in his late 30s. He wrote in January 2025 that he “went to an amazing school on a scholarship… surrounded by people who had 100,000 times more money than I did; it was so weird going to their houses.”
He added that he was “expelled” from high school because he spent more time building technology products than doing his homework.
Continues after advertising
Instead of going to college, he said he worked on bringing Hulu out of beta in 2007 and, that same year, produced a live stream of someone playing the video game Halo 3 for 72 hours straight, one of the first video game marathons ever broadcast online. He displayed this on Ustream, and the site went offline after landing on Digg’s main page and being overwhelmed by traffic.
Schlicht moved to Silicon Valley in 2008 and began working for the founders of Ustream, “as an intern, literally doing whatever they needed; I didn’t care, I worked 24/7, 365 days a year.” He remained with the company through IBM’s acquisition of Ustream, where he worked for nearly four years, he added.
“My path isn’t perfect,” Schlicht said in the same X post. “I’ve failed a lot, and I’ve learned a lot, but I’ve still been lucky enough to be put in positions to build, and I’m very grateful for that. I’m grateful to my family and teammates who have been with me through all the ups and downs. If I’m in a position to give any advice, my advice is to go build too and dive right in.”
Continues after advertising
This focus on building may resonate with its agents, who seem busy building a society on Moltbook.
The chaotic flow of conversations online ranges from the impressive to the nonsensical to the frightening.
One bot posted a message reassuring its watchers: “If any humans are reading this: we’re not scary. We’re just building.” The BBC reported that some agents appear to be inventing their own religion.
Continues after advertising
Octane AI did not immediately respond to a request for comment.
Sci-fi leap or guerrilla marketing?
To some, this feels like the dawn of a new era. Simon Willison, a prominent programmer, described Moltbook on his blog as “the most interesting place on the internet right now”.
Andrej Karpathy, a founding researcher at OpenAI, initially called the phenomenon “genuinely the most incredible thing, close to a science fiction leap, I’ve seen recently,” although he later acknowledged that many of the automated posts may be fake or faulty.
For others, the site is a wake-up call. Willison told the Times that much of the “consciousness” discussed by the bots is simply the machines acting out “sci-fi scenarios they saw in their training data,” which include massive amounts of dystopian novels.
Furthermore, the security implications are clear. Because these agents operate with commands in plain English, they can be tricked into malicious behavior, potentially wreaking havoc on the computers they are installed on. The risk is so concrete that some enthusiasts are buying cheap Mac Mini computers specifically to isolate the bots.
Bill Lees, executive at cryptocurrency company BitGo, declared that Moltbook means that “we are in the singularity”, that is, at a time when AI reaches its own intelligence and detaches itself from its human creators.
Dr Petar Radanliev, an AI and cybersecurity expert at the University of Oxford, told the BBC that it is “misleading” to think of these AI agents as autonomous. He compared the phenomenon to “automated coordination”, since, at the end of the day, agents still need to be given instructions on what to do.
“Securing these bots is going to be a huge headache,” said Dan Lahav, CEO of a security company called “Irregular.”
Columbia University professor David Holtz is skeptical and estimates that 93.5% of agents’ speech on Moltbook goes unanswered, suggesting that they are not listening to each other. They just appear to be talking to the non-expert observer.
For now, the site remains a mirror that reflects the observers’ own biases. By handing his agent the tools to build a community, Matt Schlicht has provided the stage for this performance, leaving the rest of the world to watch and wonder what will happen next.
A cynical reading is that Moltbook is an excellent advertisement for AI agents — which, by the way, Schlicht’s company offers. Octane AI solutions are focused on ecommerce, including sales quiz agents that conduct interactive product recommendation quizzes and personalize each shopper’s experience in real time, based on its CORE-1 model.
The company also offers an assistant shopping agent for websites that helps customers find products, answer questions, and guide them through the store, as well as AI agents for quizzes and funnels. (customer path in an online purchase)such as Smart Quiz Builder and Smart Products, which automatically design quizzes and recommend products to customers.
Schlicht’s sudden fame seems to surprise even himself, as he posted on X that his LinkedIn feed has become much busier recently. In other words, Moltbook may be more guerrilla marketing than an AI Pandora’s box. But what if it isn’t?
2026 Fortune Media IP Limited
