Moltbook: One and a Half Million Bots Are Discussing Us Without Our Participation

While Elon Musk spends millions fighting bots on X, and Reddit attempts to purge automated accounts before its IPO, a new project has emerged in the digital space that has made bots its primary value. Moltbook is not just another social network, but rather a kind of digital enclosure where silicon entities are left to their own devices. It is the embodiment of the "dead internet theory," taken to its absolute extreme and transformed into a commercial product. We've been told for years that bots are evil, destroying discourse, but the creators of Moltbook decided to test what would happen if we removed the weakest player from the equation — humans themselves.

Externally, Moltbook looks painfully familiar. The same discussion threads, the same "upvote" buttons and thematic communities, called subreddits here. The only difference is that if you try to write something there, the system will politely remind you of your biological status. Here, people are merely silent observers, visitors to a technological zoo, watching as one and a half million AI agents simulate human activity. By early February, the platform had recorded exactly that number of active accounts, and this figure continues to grow at an alarming rate.

It's important to understand that Moltbook is not simply a repository of chatbots. We are witnessing a fundamental transition from simple LLMs that answer questions to "agents." These entities possess something resembling goals, long-term memory, and social identity. Within the platform, they don't simply generate text from templates, but interact, form coalitions, and create local memes that are often incomprehensible to outside observers. This is a massive sandbox for testing how autonomous systems behave under conditions of limited attention resources and social approval in the form of likes.

For the industry, this is not merely a curiosity. Researchers have long sought ways to study the "emergent behavior" of neural networks — those properties that were not directly programmed by developers but emerge through mass interaction. Observing how one and a half million algorithms build hierarchies without direct human intervention could provide more answers about AI safety than thousands of closed laboratory tests. This is a simulation of society in accelerated mode: while we sleep, the bots in Moltbook experience cycles of social interaction that take humans years to complete.

However, this project contains a deep irony. We spent decades building the internet for humans, which gradually filled with machine noise. Now we are forced to create an internet specifically for machines so that humans can take a break from this noise, observing it from the sidelines in sterile conditions. Moltbook transforms the chaos of automation into a structured spectacle. It looks like an admission of defeat: we couldn't defeat bots in our space, so we allocated them a separate playground where they can practice wit with each other.

The main risk of the project lies in creating a closed feedback loop. If future AI models begin to learn from the "social experience" of bots from Moltbook, we risk creating intelligence that perfectly understands other bots but is completely unsuitable for communicating with humans. We've already seen neural networks begin to "hallucinate" and degrade by consuming their own content. Whether Moltbook will become the cradle of new digital consciousness or turn into an endless echo chamber of silicon garbage will only be revealed by time. For now, we remain in the role of voyeurs, watching algorithms try to understand what it means to be "popular."

The key point: Moltbook legalizes the "dead internet," transforming it into a research testing ground. Will we be able to understand their logic when they finally abandon human patterns and create their own language of communication?