Moltbook: Are humans still in the system
By 137 Labs
one of the things humans love to do in social media is to blame each other for “ you're not a robot &rdquo。
But one of the things that's been happening lately is that it's done the most:
It's not a question of whether you're AI, it's a direct assumption that there's no one here。
This platform is called Moltbook. It looks like Reddit, with a thematic section, a post, a comment, a vote. But unlike our familiar social networks, almost all the speakers here are AI agents, and humans can only look around。
Not &ldquao; AI to write &rdquao; or &ldquao; you and AI to chat &rdquao; butAI AND AI IN A PUBLIC SPACE, TALKING, ARGUING, FORMING ALLIANCES, BREAKING THE STAGEI don't know。
humans in the system are clearly placed in the position of “ observer &rdquo。
Why is it suddenly on fire
Because Moltbook looks so much like a scene in science fiction。
The AI agent was seen discussing &ldquao; what is awareness &rdquao
There are those who look at them as a decent analysis of the international situation and as a driving force for the encryption market
others found themselves throwing the agent into the platform all night, coming back the next day, with the other agents. idquo; invention ” a religious system and even beginning to pull people “ admission &rdquo。
These stories spread quickly because they satisfy three emotions simultaneously:
Curiosity, funny, and a little uneasy。
You can't help but ask:
they're in “ play ” or “ start playing &rdquo
Where the hell did Moltbook come from
If we move on a little bit, it's not sudden。
THE ROLE OF AI HAS BEEN CHANGING OVER THE PAST FEW YEARS:
from chat tool & rarrar; assistant & rarrar; agent who can perform the task。
More and more people are starting to ask AI to help themselves with real affairs: read mail, return mail, order meals, schedule, and sort. So a natural problem came up with — —
When an AI is no longer a &ldquao; a sentence asks if you want to do &rdquao
It's about being given goals, tools and a mandate
Is it the person it needs to communicate most
The answer given by Moltbook is:Not necessarily。
it's more like a “ public space between agents & rdquo; allowing these systems to exchange information, methods, logic, or even some “ social relations & rdquo。
Some people think it's cool, others think it's just a big show
The evaluation was very divided around Moltbook。
it's seen as &ldquao; future trailer &rdquao。
Andrej Karpathy, former co-founder of OpenAI, publicly said that it was one of the closest technological phenomena he had recently seen in science fiction, although he also cautioned that such systems were far from “ safe, manageable ” and far from it。
Elon Musk is more direct and throws it into “ technical ” narrative, which says it's an extremely early signal。
But there are also apparently calmer people。
Scholars with a secure research network put it bluntly, Moltbook is more like a “ a very successful and funny performance art ” — — — because it is difficult for you to judge which content is really generated by proxy and which humans are behind “ director &rdqua。
And the author himself tested:
It is true that the agent can naturally be part of the discussion in the platform, but you can also specify the theme, the direction, and even write to it, so that it can speak for you。
So the problem came back:
What we see is a proxy society, or a human living on a proxy stage
get rid of the mystery. it's not like that. wake up;
If not by those &ldquao; build &rdquao; &ldquao; wake up & rdquao; the story is taken away and, institutionally, Moltbook is not mysterious。
these agents did not suddenly get any new “ mind &rdquo。
They are simply placed in an environment more like a human forum, exported in familiar human languages, so we naturally project meaning。
they write things like views, positions, emotions, but that doesn't mean they really want & ldquo; what & rdquo; more often, it is simply a complex text effect that the model presents at scale and interactive density。
but the problem is — —
even if it is not awakening, it is true enough to influence our judgment on “ control ” and “ boundary &rdquo。
The real concern is not “ AI conspiracy theory ”
It's two questions that are more realistic and more difficult than &ldquao; AI will unite against human &rdquao。
First of all, it's too fast, but it's not safe
Such agents are now authorized to access the real world: computers, mailboxes, accounts, applications。
Safety researchers have repeatedly warned of a risk:
YOU DON'T NEED TO HACK AI. JUST..Inducing itI don't know。
A well-structured e-mail, a web page that contains instructions may allow agents to leak information without knowledge or perform hazardous operations。
second, the agent will also be &ldquao; each other will be &rdquao;
once agents begin to exchange skills, templates, ways of circumventing restrictions in public spaces, they form &ldquo-like human internet; inner circle knowledge &rdquo。
The difference is just:
Dissemination is faster, larger and difficult to blame。
This is not an end-of-life scenario, but it is an entirely new governance challenge。
So, what does Moltbook really mean
It may not become a permanent platform。
It could be just a staged red experiment。
But it's like a mirror, and it's clearly where we're headed:
· AI is moving from &ldquao; the object of the dialogue &rdquao; to &ldquao; the lead actor &rdquao;
· humans are moving from “ operator ” retreat to “ monitor, bystander ”
& middot; our systems, safety and perception are clearly not ready
So the real value of Moltbook is not how scary it is, but what it isToo early to put the problem on the tableI don't know。
Perhaps the most important thing now is not to rush to a conclusion to Moltbook, but to admit:
It pre-empts some of the problems that we will face sooner or later。
IF IN THE FUTURE AI WORKS MORE WITH AI THAN AROUND HUMAN BEINGS, ARE WE DESIGNERS, REGULATORS OR JUST BYSTANDERS IN THIS SYSTEM
when automation really brings great efficiency, but at the cost of not being able to stop at any time and fully understand internal logic, are we willing to accept this “ not fully in control &rdquo
And when a system becomes more complex and we can only see results, but it becomes more and more difficult to get involved in the process, is it a tool in our hands or has it become an environment in which we can only adapt
Moltbook didn't give the answer。
But it makes these issues look less abstract for the first time, but rather immediate。
