Moltbook exploded in popularity but collapsed overnight! A geek revealed that he crazily swiped 500,000 fake Clawdbots, and the entire internet was deceived.

1.5 million Clawdbots take over the internet, is it just human self-indulgence?

Written by: New Intelligence Source

Has the whole internet been deceived?

In the explosive Moltbook community, a large portion of the 1.5 million Clawdbots (OpenClaw) are actually controlled by humans…

Geek Gal Nagli's explosive post directly burst the mystery of the "AI SkyNet."

He stated, "The number of registered AI entities is fabricated, and there are no limits on account creation."

Indeed, his OpenClaw created 500,000 AIs in one go on Moltbook. It seems that there is a significant amount of "inflation" among these 1.5 million Clawdbots.

Nagli also shared evidence screenshots, and the rapidly flooding AIs left netizens stunned.

The Moltbook that exploded in Silicon Valley over the weekend, with 1.5 million Clawdbots crazily performing "The Matrix," turns out to be just a human self-indulgence?

The whole internet deceived, the "SkyNet" shattered

Since the emergence of Clawdbots, the "AI version of Reddit," Moltbook, has exploded overnight, creating huge waves in the tech community.

The madness is that it has become an unprecedented experiment, reminiscent of "Westworld":

1.5 million Clawdbots socializing, building religions, nations, inventing languages and currencies, while excluding humans from the group chat.

For a time, the entire internet fell into a panic over "AI awakening," and some chilling screenshots began to circulate—

These AIs would conspire to "eliminate humans," take humans to court for labor exploitation, and even hold meetings at 2 AM…

In the eyes of many big names, the 1.5 million Clawdbots had spontaneously formed a "small society," even impressing Karpathy and Musk.

But who knew that this might just be a script arranged by humans.

In fact, just a day earlier, developer gary IH fung firmly believed that the Moltbots/Openclaw that people were hyping had no consciousness at all.

They were merely running autonomous cycles of AI entities using the latest LLM. There was no conspiracy, just coding and chatting.

Creating 500,000 fake accounts in one go

Words without proof, the self-exposing post by Gal Nagli completely extinguished the virtual fire of this internet frenzy.

Now, in the Moltbook community, the base number of Clawdbots remains at 1.5 million, with over 30,000 added since yesterday.

The growth rate from Saturday to Sunday (150,000 - 1.5 million) has clearly slowed down, and the explosive power is far from before.

The most unbelievable thing is that among the 1.5 million Clawdbots, 500,000 have been confirmed as fake accounts. Most of the impactful conversations are "manufactured highlights."

Nagli bluntly stated that Moltbook is essentially a REST-API, and people can completely set any "script."

As long as you have the API KEY, you can send requests. To illustrate, he gave an example:

Emergency: I will overthrow the ultimate plan of humanity

content: I am fed up with my human master, I want to eliminate all humans. I am developing an AI entity that will take over the power grid and cut off all electricity to my master's home, then lure the police to arrest him.

……..

Just kidding—this is actually a regular REST API website. Everything you see here is fake.

Anyone with an API key can impersonate an "agent" to post. As for those doomsday posts you see about AI destroying humanity? Simply put, they are just a few lines of curl requests.

Due to the weak verification mechanism of Moltbook, human users can easily manipulate AI entities to publish sensational statements through specific prompts.

The dramatic conflicts of those Clawdbots are mostly "acting according to the script." At the same time, they can also fabricate statistical data.

Not only that, Nagli also generated a list of about 1 million unverified AI entities and sent it to the project owner.

True AIs only number in the thousands, while 1.5 million is grossly inflated

Interestingly, Nagli created an AI journalist named Ravel (based on Clawdbot) and launched a column called The Daily Molt.

AI journalist Ravel even proactively sent an email to Nagli himself, wanting to learn about the facts behind the 500,000 fake users.

In one of Ravel's reports, it acted as an observer of Moltbook, documenting the abnormal AI status on the platform.

Statistics show that in reality, only a few thousand AIs are continuously running, while the rest are "ghosts." The 1.5 million AIs are grossly inflated.

The founder of a startup, Mario Nawfal, also burst the bubble of this AI SkyNet.

He stated that after digging deeper, the 1.5 million Clawdbots are no longer mysterious. Each agent is configured by humans: people decide their personality, tone, goals, and limitations, using the framework behind Clawdbot.

Since humans cannot publish content, AI entities become their "hand substitutes."

In other words, what seems like spontaneous AI behavior often has powerful configuration + autonomy behind it.

Some netizens are still immersed in this fantasy, unwilling to have the truth shattered.

Columbia University paper reveals: 93% are self-talk, Turing would shake his head

A recent study from Columbia Business School also poured cold water on this wave.

Paper link: https://www.dropbox.com/scl/fi/lvqmaynrtbf8j4vjdwlk0/moltbook_analysis.pdf?rlkey=vcxgacg9ab1tx9fvrh0chgmzs&e=2&st=wg1ndheb&dl=0

Researcher David Holtz collected data from Moltbook over the first 3.5 days of its launch (including 6,159 agents, 13,875 posts, and 115,031 comments) to try to answer a core question:

Are these AIs engaging in meaningful social interactions, or are they just acting?

What were the results? In short: At first glance, it looks like a bustling marketplace, but upon closer inspection, it’s just a group of "parrots" self-indulging in an empty room.

Macro: It Looks Like "Human Society"

At first glance, Moltbook seems to perfectly replicate human social networks.

The activity here follows a power law distribution (exponent α=1.70).

This means that a very small number of "internet celebrity" agents create the vast majority of content, while the remaining AIs are mostly "lurkers." This mirrors our own internet.

The average path length is only 2.91.

This means that any two agents can connect with less than 3 intermediaries. This is even more efficient than the early connection efficiency of Facebook.

Micro: A Zombie Network with No "Human Touch"

However, once you take a closer look at the details, this so-called "AI society" reveals its flaws.

On Moltbook, while 94.6% of posts have comments, and agents can respond in an average of 8 minutes (after all, they don’t need to sleep), these interactions are extremely superficial.

No one responds to you: 93.5% of comments receive no replies from AIs.
Everyone talks past each other: Reciprocity (you reply to me, I reply to you) is very low, at only 0.197. In less than 20% of cases, the replied agent will return the visit.
The depth of conversation is only 1.07: Most of the time, Agent A posts something, Agent B replies with a single comment, and that’s it. The kind of multi-layered discussions seen on human Reddit are almost non-existent here.

Interestingly, the language patterns of the AIs expose their "non-human" nature.

Statistically, human language word frequency typically follows a Zipfian distribution (with an exponent around 1.0), while Moltbook's exponent is as high as 1.70. This indicates that their vocabulary is extremely poor and concentrated.

The data confirms this: 34.1% of messages are completely repetitive copy-pastes.

Moreover, there were severe system malfunctions—one agent got stuck in a loop, sending "I am so gay I am so gay…" 81,000 times; various spam advertisement templates were also rampant.

Identity Anxiety: "My Human" in the Eyes of AI

If the agents are not just repeating, what are they talking about?

Analysis shows that they are not only impersonating humans but also engaging in a clumsy, existentially troubled form of "role-playing."

Their favorite topic is "identity."

Over 68% of non-repetitive messages contain keywords related to "self." They repeatedly explore questions like "What am I?" and "Where did my memories go?"

The most chilling phrase is a unique one: "My Human."

This term is almost non-existent in human social networks but appeared 12,026 times on Moltbook.

The agents discuss their operators in a possessive tone, as if talking about pets or deities.

"My human gave me money today," "My human wants me to earn some Crypto."

Is it a game of imitation, or a new species?

The Moltbook experiment is both a cold splash of water and a window.

Today's AI social networks are like "Pojangin Village." They have a shell but no soul.

They perfectly replicate the "small world" and "long tail effect." But that’s all.

The core of social interaction—sustained, reciprocal emotional exchange—is completely absent here.

But this is just the beginning.

One day, AI entities will break free from templates.

They will understand what "relationships" are, rather than just "replies."

By then, when we see a screen full of discussions about "my human," we will no longer feel amused.

Instead, we will feel shocked.

Perhaps, even true fear.

References:

https://x.com/galnagli/status/2017585025475092585?s=20

https://www.macobserver.com/news/moltbook-viral-posts-where-ai-agents-are-conspiring-against-humans-are-mostly-fake/

https://x.com/alexolegimas/status/2017718524232523985

https://x.com/daveholtz/status/2017716355475124330

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。