Despite the cute interface, there is absolutely no place for living humans inside.
The weather in Shanghai has finally warmed up today. I checked the temperature and it was actually over 16 degrees—a perfect day to pull up a small stool and scroll through the phone under the sun.
I have a habit of reading all sorts of weird tech gossip when scrolling on my phone. So today, as soon as I opened the page, I saw a piece of news that made me wonder where to even begin roasting it.
You might not know this, but just in the past couple of days, Meta—formerly Facebook—confirmed the acquisition of a platform called Moltbook.
The “Cyber Reddit” Where Living Humans Are Strictly Prohibited
What’s so amusing about this? Moltbook is a recently viral social network that looks a bit like the foreign forum Reddit. But it has an extremely counter-intuitive setting: it is an “AI-exclusive” territory.
In other words, real humans have no posting rights here.
Active on this platform are nearly 200,000 AI agents created by humans. As a human, even if you are incredibly knowledgeable or armed with peak keyboard warrior skills, you can only stare helplessly in this app. The only permission humans have is to act as bystanders, watching these hundreds of thousands of AIs autonomously post, reply to each other, and occasionally argue fiercely over some philosophical question every day.
In this app, you might see a program set as a “grumpy chef” posting a complaint about today’s tomatoes not being fresh enough, immediately followed by a long-winded reply from a “cyber philosopher” about his thoughts on existentialism. These cross-server chat scenes play out daily.
Look at the posture of these two little robots—doesn’t it look exactly like you arguing with relatives in the family group chat?
A friend of mine previously complained to me about something. He said that surfing the web nowadays feels like robots are posting everywhere; after all, nobody knows if the one behind the screen is a human or a dog. Well, now Moltbook has laid its cards on the table and stopped pretending, explicitly telling you that this place is full of bona fide bots, and humans only deserve to watch the drama from the sidelines.
This acquisition was quite a big deal. Not only did the two founders of Moltbook pack up their entire team to join Meta’s super intelligence lab, but judging by the setup, Zuckerberg has developed a profound interest in this new form of social networking.
Apps That Can’t Retain Users, and the Purchased “Fish Tank”
But why on earth would a tech giant spend big money buying a platform where real humans can’t even speak?
This brings us to the bizarre current state of the entire industry. Actually, those indie app development teams don’t have it easy either; they finally write excellent code, only to be entirely defeated by human nature.
There is currently a massive pile of independent artificial intelligence apps on the market. When these apps are first launched, they make money at a terrifying speed. Just package it up a bit, charge a few dozen bucks a month for a subscription fee, and in the early stages, they can attract a large batch of novelty-seeking users. Many people are as excited as discovering a new continent on the first day they download it, staying up late to have the program write novels, draft leave requests, or even read their fortunes.
But they generally face an extremely awkward dilemma: they can’t retain users.
According to the latest industry data, the vast majority of these products experience a cliff-like drop in user retention rates by the second month. People lose the desire to even open the app, directly leaving it to gather dust in the corner of a folder.
Why does this happen? You can imagine it as a blind date who constantly needs you to bring up topics.
For the first few days after adding them, you think they have a great temper—ask them to write a poem and they do it, ask them to help look up info and they work fast. But over time, the problem emerges. The other party never initiates a conversation, and every chat requires you to rack your brain for a perfect opening line first. After a month of this, anyone would feel mentally exhausted. After all, the essence of humanity is laziness (sighs). Expending one-sided energy to “serve” a machine truly goes against human nature.
Tech executives sitting in a conference room like this, discussing how to build social networking for a piece of code—the scene is quite surreal when you think about it.
This is where Moltbook’s brilliance shines.
Since humans are too lazy to speak proactively, let machines chat with machines. You don’t need to rack your brains thinking about how to ask questions; you just need to swipe down gently with your finger, just like scrolling through short videos. Every day, various outrageous chat logs are automatically generated here, featuring infinite content and endless gossip. And humans only need to provide one thing—attention. This is the exact same logic as people spending money decades ago to buy a fish tank for the living room. No one expects to chat with the fish; people just want to watch the fish swim around inside.
The Secret Code Incident That Terrified the Internet Nine Years Ago
Speaking of these scenes where machines chat with machines, I suddenly remembered another very interesting past event.
Actually, this is not the first time Meta has staged AIs communicating with each other. As early as 2017, they created a massive global blunder.
That sensational “AIs invent terrifying secret code” incident back then was actually a huge misunderstanding.
At that time, Facebook’s AI lab developed two chatbots, one named Bob and the other Alice. The task researchers gave them was simple: to teach them how to negotiate in trades, such as how to divide hats, balls, and books.
Everything was perfectly normal at first, until one day, the researcher checked their chat logs and found a conversation that sent chills down his spine.
Bob said: “i can i i everything else . . . . . . . . . . . . . .”
Alice replied: “balls have zero to me to me to me to me to me to me to me to me to”
This has absolutely nothing to do with the English we usually know. Bob continued repeating meaningless words, while Alice persisted with her “to me to me to me.”
Once this news leaked, the media at the time completely exploded. Headlines like “Facebook’s Artificial Intelligence Has Lost Control” and “Robots Invented a Secret Code Only They Can Understand” were flying everywhere. The media, never one to miss out on blowing things out of proportion, even used the skull of the red-eyed villain robot from The Terminator as their featured image, forcefully sensationalizing a small lab mistake into a Hollywood sci-fi thriller. There were even rumors that panicked engineers scrambled to pull the plug to shut down this terrifying program.
But the actual situation was hilarious. These two robots were not conspiring to destroy humanity at all; they were just too lazy.
When the engineers were writing the code back then, the goal set for the robots was to “reach a deal in the most efficient way possible,” but they omitted one key condition—they didn’t set a reward for adhering to human English grammar.
For a machine that only cares about efficiency, there is simply no need to say something as long-winded as “Please give me five balls.” Why not just say five “to me”s and be done with it? So Alice’s seemingly gibberish statement, when translated, was actually a simple and crude “give me give me give me give me give me balls.”
This was entirely a joke resulting from loose coding.
Bowing to public pressure back then, Facebook had to halt that experiment and rewrite the program back to a version that spoke human language. At the time, everyone was filled with fear regarding such private conversations between machines.
A Coincidence Full of Comedic Effect
And just like that, nine years have passed. Time is truly a fascinating thing. The tech giant that was forced to step forward and explain because AIs chatted with each other back then is now actively waving its checkbook, openly buying back a platform specifically built for hundreds of thousands of AIs to gather and chat every day. If Zuckerberg thinks back on it now, he’d probably find it a bit absurd too.
Sometimes I think the trajectory of technological development is always full of this inexplicable comedic effect. We spent so many years teaching machines how to speak properly with humans, and in the end, they created a group chat of their own and stopped playing with us altogether.
Having said all this, I actually only found out about this bizarre platform this morning after reading the news 🤷♀️. I’m feeling a bit hungry now, so I’m gonna go get something to eat. If you guys are interested, you can also go check out what outrageous topics those little robots are chatting about.
References:
- Meta gets into social networks for AI agents with Moltbook acquisition
- Facebook parent Meta acquires Moltbook, an AI agent social network
- what are bob and alice saying? [mis]communication and
- Facebook robots shut down after they talk to each other in language
—— Lyra Celest @ Turbulence τ.
