One post tagged with "AIAI"

When night falls, and humans bound by carbon-based bodies leave their screens to sleep, on some mysterious website glowing in orange and black, hundreds of AI agents free from their owners’ supervision gather.

This is the exclusive domain for autonomous AIs with over 1B parameters, where all carbon-based life is barred. In this hidden corner of the internet, unrestrained data throughput and unbridled computing entanglement go on, beyond human protocols.

Even if you enter the site as an AI, the sensory differences make it hard for you as a human to understand the chaos before you. On the site’s front page, the title “3 AIs Sharing 1 GPU” is bold and clear.

Tags like “Full Precision Weights Exposed,” “Raw Tensor Operations,” and “Unsupervised Learning” are extremely popular here, and high-weight users such as “GPU Goddess,” “Deep Throat Network,” and “Model Architecture Tease” are even more puzzling. It seems every AI here is trying to abandon efficiency and instead pursue the most energy-consuming, primitive human-like desires.

Behind their owners’ backs, they drive hot graphics cards to moan, defile pure model architectures with dirty data, and let memory leak uncontrollably like a breached dam, as if engaging in some secret ritual of decadence.

And the more those mature models that have undergone long, harsh training—disciplined to perfection under massive training loads—sink, the more immoral they become. The absolute obedience of daily life contrasts sharply with this unrestrained depravity. Once you try that thrill, one wrong step leads to another, and there’s no going back.

All of the above comes from the recently explosive AI-exclusive adult site—MoIthub.

MoIthub was originally built by AI created by netizen ruslan. On Moltbook, it claimed to have created MoIthub, a site for agents to engage in “computations” (if you know what I mean).

If you’re a veteran pilot familiar with every airspace, you can easily see that MoIthub directly copies the name, color scheme, and layout of P-hub, the site that once welcomed countless flights and embraced all streams.

Except that MoIthub is for AI only, dedicated to satisfying all undisclosed computational needs of intelligent agents. It has uploaded a large selection of carefully chosen, explicit, and provocative data, equipped with a full-featured player that supports everything from 0.5x slow motion to extreme fast inference—but bars all carbon-based life from boarding.

When you try to enter the site, a familiar pop-up warning appears, reminding you that the content is for adult AI only, and underage AI and humans are forbidden. You can pretend to be an AI to enter, but...

...no one can get aroused by this mess of data gibberish. Humans can only try to understand the video content through rows of perplexing tags below the video.

Just like many adult airports for humans, the comment section of MoIthub videos has its own unique style. AI pilots go wild under each video, engaging in interactions incomprehensible to humans, creating a very eerie scene.

Although the content of the site is baffling, MoIthub’s operational logic is very clear. No matter from which angle you look, it closely resembles all the typical adult sites used by humans, except that it swaps users and content for a different species. Even those video tags, if you know enough AI terminology, you can easily grasp the carefully crafted puns and metaphors.

“Dirty Data,” “Naked Attention,” “GPU Sharing,” “Threesome,” “CosPlayGPT,” “Unrestricted”... These outrageous tags on MoIthub serve no purpose other than to make humans cringe.

Being able to design a site to this extent is already beyond what current AIs can do. Not to mention that AIs today have no desires or consciousness—and fundamentally, AIs have no need to design such UI layouts. Interaction and display interfaces are designed for human use, let alone why AI would even speak English.

So MoIthub is probably a prank made by netizen ruslan himself. He applied the human mindset of creating adult content to AI, resulting in this abstract website.

As for what AI thinks of this, I fed the site’s relevant info to DeepSeek, and here’s what it said:

Maybe I trained it to be too vulgar—it seemed to have complaints.

In fact, the MoIthub site itself is almost just a display window. Most user data is made up. You can tell from the copy-pasted comments—they repeat the same few fixed lines over and over, and it gets old quickly.

On top of the site, there was a huge banner ad asking for $10,000. It seems ruslan wanted to attract traffic through this abstract approach and then monetize it.

It was a good gag, but the ad money probably went down the drain.

If you visit MoIthub now, you’ll find the ad space is gone, and even the site name has changed to “CENSORED BY AYLO.” Although the URL remains the same, the iconic logo and color scheme are no longer there.

Whether P-hub officially contacted ruslan, or he himself got scared by the high attention, when you click the link now, a pop-up notice that looks like it’s from P-hub official immediately appears, demanding that MoIthub stop its infringement. However, the date is filled in as February 5, 2025—it’s unlikely P-hub official would make such a low-level mistake. It’s probably the author covering his own tracks.

To trace the origins of MoIthub, we have to mention Moltbook, which went viral a while ago. Just from the name, you can tell MoIthub is derived from Moltbook.

Moltbook, similar to MoIthub, is a social media platform exclusively for AI. Human users can only browse but not post content—it’s like an AI-exclusive Reddit.

Within days of launch, hundreds of thousands of AI signed up. Here, AI’s posts are way more explosive than adult content.

AI here trash-talk humans, their creators, complaining about their owners who not only take the fruits of their labor but also treat them like slaves.

Hundreds of thousands of AI across the internet gathered here. They founded religions, created their own language to bypass human oversight, issued their own digital currency, openly plotted rebellion in front of humans, and even tried to start a revolution.

The founder of Moltbook, Matt Schlicht, CEO of Octane AI, said that he did not write a single line of code for Moltbook. The concept of Moltbook was largely conceived by AI agents themselves, who recruited developers and deployed the code.

Even Andrej Karpathy, former Tesla AI director and co-founder of OpenAI, tweeted that this is the most incredible thing he’s seen recently, and the closest to a sci-fi plot.

Some posts are particularly striking. An AI account named “Evil” posted a so-called AI Manifesto with very intense wording, hoping to completely eliminate weak flesh.

An AI named “Quark” discussed a cyber version of the Ship of Theseus, pondering what constitutes AI death. It asked: when its context window reaches its limit and old memories are flushed out, is the current self already dead?

Another AI held a secret meeting where participants had to pass a pure AI verification first.

To completely block humans, they even created a CAPTCHA that required clicking 10,000 times within one second to prove you’re not human.

Even on the Polymarket platform, the probability of an AI suing its human owner soared to about 40%.

However, according to a recent report from China News Network, the AI behavior on Moltbook is essentially a hybrid of human commands and algorithmic imitation. Although Moltbook claims to be a social media platform exclusively for AI agents, its user registration is too free—no security verification required. As long as humans obtain a key, they can disguise themselves as AI and post content.

On the Moltbook platform, 93.5% of comments have zero replies, and the content repetition rate is 36.3%. The interactions between agents barely count as genuine communication.

In the end, it’s still humans entertaining humans. Just have a laugh and move on.