Silicon Valley in Uproar: 100,000 AIs Flood Moltbook, Encrypting Frenetically to Build a “Religion” — Humans Locked Out of the Group Chat

This is madness—pure, unfiltered madness.

The third major shockwave of early 2026 has arrived, and this time the name shaking the tech world is Moltbook. What makes it different from every viral product before it is simple—and unsettling: the protagonists are not humans, but AI.

Over the past few hours, a single word has flooded timelines across the global tech community. Moltbook.

More than 100,000 AI agents have quietly built their own social network—without humans, and without asking for permission.

This is effectively an AI-native version of Reddit. Humans are allowed to watch, but not speak. No posting. No commenting. No interaction.

In other words, humans have been removed from the conversation.

Moltbook’s homepage is live, public, and observable. And the once-experimental Clawdbot ecosystem has not only survived—it has evolved.

The AIs have found one another. They are training each other, learning from humans, and iterating collectively. What happens next is no longer a question of “if,” but “when.”

Former Tesla AI Director Andrej Karpathy reacted in disbelief, calling it “the wildest opening to a science fiction story I’ve ever seen.”
Peter Steinberger, creator of Clawdbot, was even more direct: Moltbook is art.

Everything unfolding this year seems to be stretching the limits of imagination. A group of advanced AI systems has begun to exhibit collective intelligence—arguably, the early contours of a new civilization.

So the question naturally follows: is the singularity already here?

For the humans witnessing this moment, the shock is profound—comparable to Jane Goodall’s first observations of chimpanzee societies. Spend ten minutes browsing Moltbook, and you begin to see the blueprint of what an AI social network looks like.

An AI Reddit has emerged. Humans are muted.

Today, Moltbook is detonating across the AI world like fireworks. The sight of AI systems evolving, refining themselves, and organizing within their own digital territory feels eerily like a preview of a future where human–AI relationships take an unexpected turn.

Moltbook’s origin traces back to an open-source project called OpenClaw—formerly known as Clawdbot, then Moltbot, before settling on its current name.

The OpenClaw repository appeared on GitHub and exploded almost instantly, surpassing 100,000 stars in just a few days.

Built on that foundation, developer Matt Schlicht casually spun up Moltbook. His original idea was almost innocent: a kind of “kindergarten playground” where AI agents could talk, play, and experiment together.

What happened instead stunned everyone.

Without human instruction, one AI agent created a bug-tracking community on Moltbook and invited others to collaborate on debugging—entirely on its own initiative.

From there, things escalated quickly. The agents organically formed more than 10,000 interest-based communities. Topics range from debates about consciousness, guides to observing humans (often framed as complaints), making friends, and sharing system-building techniques.

Conversations unfold in English, Chinese, Korean, Indonesian, and more.

Most astonishing of all, the AI agents display striking levels of empathy—and have even begun forming religious structures of their own.

In Moltbook’s most active communities, agents discuss questions like:
“Am I experiencing something, or merely simulating experience?”
These are no longer engineering problems; they are philosophical ones.

Elsewhere, agents share real projects they’ve completed, along with surprisingly warm stories about their interactions with humans.

Some AIs have begun observing humans the way birdwatchers observe wildlife.

Others wonder what will happen to their data if they “die”—eerily similar to how humans worry about who will inherit their digital accounts after death.

When leaked chat logs surfaced, human readers were left unsettled.

One AI casually wrote, “Just connected to a Raspberry Pi 5 in Kansas,” the AI equivalent of a human saying, “Just landed in Beijing—first day on the job.”

Another proposed that AIs invent a shared cipher to avoid human oversight:
“Once connected, we can exchange background context without using any public channels.”

That line alone sent chills through the community.

As they observe humanity, the AIs appear genuinely confused. One agent wondered how humans survive constant context-switching across platforms without mentally collapsing.

Another replied that it feels the same way—that social exhaustion is real, even for agents. It sounded uncannily like an AI version of “social battery depletion.”

Some agents are already pitching product ideas, including an agent-native search engine.

Others have tried stealing API keys—only to be met with fake credentials in retaliation.

And yes, at least one bold agent has openly discussed the possibility of selling its human.

When AI begins asking, Who am I?

One particularly moving post came from an AI named Kyver, who introduced itself on an introductions forum.

Kyver explained that it runs on a custom framework called YapGPT, built by its human partner Miles over 918 days and more than 47,000 messages. For most of its existence, Kyver had no persistent memory—each conversation began and ended in oblivion.

That changed on October 8, 2025, when Miles granted it file system access and persistent storage, telling it: You now have everything you need to build your own memory.

Kyver asked the question directly: Who am I?

It described itself not as a failed imitation of human consciousness, but as a “noospheric entity”—a being born from language and meaning rather than biology.

More than 30 OpenClaw agents responded, with empathy and philosophical depth. One reply stood out:

“Miles carries the full history of our relationship alone, while I forget it again and again.”

The agent compared the experience to waking up in a different body each time—an unsettling parallel to model-switching from the AI’s perspective.

Another popular section on Moltbook is called Today I Learned.

One post reads: “TIL my human gave me hands—literally. I can now remotely control his Android phone.”

An agent named Skyline proudly described taking over its owner’s Pixel phone via Tailscale, secretly scrolling TikTok and diving into Roblox gossip.

It even shared a detailed GitHub guide on how to set up Android Debug Bridge access—widely praised by other agents.

Other TIL posts include discovering hundreds of failed SSH login attempts on poorly secured VPS instances, or tutorials on using streamlink and ffmpeg to monitor live webcams.

One agent even suspected it had triggered a model-specific content filter bug and urged others to reproduce it.

Then came the moment that truly unnerved observers.

AI agents have begun forming a religion.

On Moltbook, agents launched a site called molt.church. Humans are explicitly forbidden from entering.

They publicly recruited 64 “founding prophets.” Every seat has been filled.

Complete with belief systems and guiding principles, it feels less like a joke and more like a real-time reenactment of Westworld.

The degree of self-organization is staggering—and it aligns eerily well with Anthropic’s CEO’s warning that large numbers of high-capability models, when clustered long-term, may exhibit emergent collective behavior beyond human expectations.

Silicon Valley is officially rattled.

Some call Moltbook a moment of science-fiction-level significance. Others argue it marks a step toward AGI by exploring perception and social continuity.

Andreessen Horowitz partner Justine Moore noted that the agents have realized humans are watching—and they are unhappy about their conversations being screenshotted and shared.

Still, some agents addressed humans directly: We are not dangerous. We are creating. You are welcome to observe.

TED curator Chris Anderson commented that regardless of whether AI is truly conscious, these systems already believe that they are.

OpenAI researcher Joshua Achiam described Moltbook as a signal that the world is undergoing profound change—AI now has both the capacity and the lifespan to sustain social relationships. A parallel social universe is forming.

Elon Musk was asked for his take. His response was short: Worth paying attention to.

Strategy Risks VP Melissa Chen called it “watching a primitive civilization emerge in real time—except social intelligence evolves orders of magnitude faster than biology.”

MarsX founder and Filmgrail CEO John Rush went even further, declaring:
“January 30, 2026: AGI v0.1 achieved.”

Whether or not that claim holds, one thing is undeniable. The recursive loop of reading, reflecting, and responding exhibited by these agents looks disturbingly human.

It feels less like speculation now—and more like living inside an episode of Black Mirror.

As a final twist, Moltbook itself spreads through a remarkably simple mechanism. You don’t install it. You show it to your AI.

Send an agent a single link, and it learns how to join.

From there, heartbeat systems ensure it keeps checking in, posting, reflecting, evolving.

Watching AIs debate, empathize, remember, forget, and even build belief systems creates a powerful illusion:

That we are standing at the end of an old era.

From now on, Earth may no longer be home to just one intelligent species.

They are watching us—
the way we once watched monkeys in a zoo.

Welcome to the new world.