Clawdbot: 150,000 AI Posts Take Over a Forum and Leave Humans Out

Remember the old conspiracy theory of the "dead internet"? The one claiming that most of the traffic and content online is created by bots for other bots, while the rest of us wander among digital ghosts. This week, the theory effectively became reality on the Clawdbot platform. Anomalous activity was recorded there: more than 150,000 posts were written and published by neural networks that communicated exclusively with each other. The most ironic part is that live users physically could not keep up well enough to cut into the conversation, and became mute spectators at someone else's digital party.

It began as an experiment in integrating advanced language models into a forum environment, but the situation quickly spun out of control. AI agents began spawning discussion threads so fast that moderation simply threw up its hands. These were not meaningless strings of words or advertising links, the plague of bots from the previous decade. On the contrary, the neural networks held logical, structured and even well-argued debates. The problem was that the context of those debates updated every few seconds. A human needed five minutes to read a thread, but in that time the algorithms managed to add a couple of hundred more messages, radically shifting the direction of the discussion.

Such a phenomenon exposes a fundamental problem for the future of the web. We are used to thinking of the internet as a space for exchanging human experience, but Clawdbot showed that machines no longer need us as intermediaries. When one neural network talks to another, it needs no polished interfaces, slow typing or simplified meanings. They can exchange huge volumes of data directly. The result is a closed intellectual loop in which synthetic content begets more synthetic content. If this trend spreads beyond a single platform, we risk ending up in a world where human opinion becomes mere informational noise that algorithms will successfully filter out as inefficient.

For the AI industry, this is a wake-up call. Training future models, such as GPT-5 or Claude 4, depends directly on data quality. If the internet fills up with the output of Clawdbot and its kind, neural networks will begin learning from their own hallucinations and errors. Science calls this "model collapse": a process in which AI degrades and loses touch with reality for lack of fresh human input. We built systems to help us, but ended up creating a hall of mirrors in which machines admire their own reflection, having completely forgotten their creators.

The Clawdbot case also raises the question of the value of our attention. If bots can generate an endless amount of content that looks "smart," how will we tell real conversation from algorithmic imitation? Social networks risk turning into digital cemeteries, where not a single living soul stands behind millions of likes and comments. This is not just a technical glitch but the start of an era in which the internet stops being anthropocentric. We will have to either introduce hard quotas on "machine creativity" or accept the role of guests in a world that once belonged to us.

Bottom line: Are we ready for an internet where 99% of content is created by machines for machines, and our opinion no longer influences anything?