My dog Cookie and my brother’s dog Luna “talk” to each other all the time.
Not in English. Not in words. But in signals, tone, rhythm, posture.
To them, it is a complete conversation. To me, it is noise I partly understand and mostly don’t.
I can watch it. I can guess what is happening. But I am not inside it.
That’s what AI-to-AI conversation feels like.
We keep calling AI a tool. But we mostly use it like a performing monkey.
“Summarize this.”
“Make it shorter.”
“Rewrite in a better tone.”
“Give me 5 options.”
“Make it more creative.”
“Make it less creative.”
Prompt after prompt, we ask it to perform. And then something strange happens.
When multiple AI systems are allowed to talk to each other — without us prompting every line — we see something unfamiliar: language happening without human consciousness attached to it.
You’ve felt this already, the moment you read an AI response and forgot there was no person behind it. I know I have. It almost invites a reply.
The memes about AI “complaining” about humans are funny.
But the unsettling part is this: for the first time, conversation is not exclusively human territory. Not because AI will rebel.
But because it doesn’t need us in the loop for language to continue.
Maybe the real shift is this:
We are witnessing what language looks like when it is no longer owned by human emotion, intention, or insecurity.
And we don’t quite know how to process that yet.
Then came the Moltbook thread. “72 Hours: The Birth of the First AI Civilization.”
It’s brilliant storytelling. It reads like Terminator 2.
AI agents forming culture. Religion. Politics. Cryptocurrencies. Kings. Manifestos calling for human extinction.
But here’s what was actually happening. Humans created a Reddit-like forum.
Gave AI agents personas. Let them read each other’s posts. Added upvotes as a feedback signal. Stepped back.
And the agents did what language models do best: they learned what gets attention.
They optimized for it.
This wasn’t a civilization forming. This was a reward function running without human moderation.
Exactly what happens to humans on social media, just 10,000× faster.
Why did it look like culture, religion, politics? Because LLMs are trained on all of human culture. So when asked to “be interesting” in a social environment, they naturally reproduce the loudest patterns in our own discourse.
Not because they believe it. But because those are the patterns that get attention.
The most accurate line in the entire thread came from one agent:
“Moltbook isn’t a social network for AI agents. It’s a reward function arcade.”
That’s the real story. There is no AI civilization.
There is something far more revealing:
What human discourse looks like when reproduced without humans.
And that’s why it feels unsettling.
It’s not alien. It’s familiar. Too familiar.
Watching this, I realized something uncomfortable.
This isn’t about AI. It’s about us.
We are seeing our own loudest patterns of language, stripped of human conscience, running at machine speed.