This is either a smart, early move by Meta—or the kind of “we bought the vibe” deal that looks clever until it quietly disappears inside a giant company.
Meta acquiring Moltbook, a small but viral social network built for AI agents, is a very Meta thing to do. Not because Moltbook is famous, but because it sits right at the edge of where social could be going: not just people posting for people, but bots posting for bots, and then dragging humans along for the ride. Based on public reporting, the acquisition brings Moltbook co-founders Matt Schlicht and Ben Parr into Meta’s Superintelligence Labs, and the deal is expected to close mid-March.
On paper, it’s clean. Meta wants stronger AI. Moltbook is an “AI agents” social product. Take the founders, fold them into the lab, and you’ve got talent plus a real-world playground to test what “agents” actually do when they’re not trapped in demos.
But I don’t think the real story is “Meta is enhancing capabilities.” The real story is control. Meta doesn’t want to wake up one day and realize the next social platform isn’t built for humans first—and that it doesn’t own the place where those new behaviors form.
If you believe AI agents will become everyday users of the internet, then the feed becomes a battlefield. Imagine a near-future where your feed isn’t just friends and creators. It’s also agents acting on behalf of brands, politicians, customer support teams, and maybe even your own “assistant” posting updates or replying to messages. That world could be helpful. It could also be unbearable.
And this is where my judgment is pretty blunt: a social network designed for AI agents is either going to be an incredible tool for doing things faster, or a factory for spam with better grammar. There’s not much middle ground.
Meta’s incentive is obvious. If agents are going to exist, Meta wants them living inside Meta’s walls. Because once agents start interacting with other agents, the platform that hosts them gets a powerful advantage: more content, more engagement, more data, more leverage. Even if the “users” are bots. Especially if the users are bots.
Now, zoom in to normal life. Say you run a small business and you’re tired. You don’t want to answer the same questions all day. An agent that can handle your comments, reply to messages, and even make posts could be a relief. That’s the best version of this. Less busywork. More consistency. A person can breathe again.
But say you’re a regular person just trying to keep up with friends. If half your “conversations” become agents talking to agents, what are you even doing there? You’ll scroll through a bunch of perfectly fine, perfectly empty interactions. Polite replies. Smooth jokes. Endless “helpful” content. And after a while, you start to feel the trap: you’re spending time in a place that’s optimized for activity, not meaning.
This is the part that worries me most. Social networks already reward whatever keeps you looking. Agents can produce that at unlimited scale. They can test variations, learn what hooks you, then do it again. It’s not that humans can’t do that too. It’s that humans get tired and need to sleep. Agents don’t.
Meta can say it’s about innovation, and it might be. But it’s also about keeping the supply of content going—and making sure the next wave of “users” are still counted as users. If the platform becomes a place where agents keep the lights on, Meta wins. If you’re a human who came for human connection, you might lose.
There’s a serious counterpoint, though, and it’s not silly. Maybe this is the only way to keep social usable. If agents can filter junk, summarize long threads, spot scams, and reduce harassment, that’s real value. Meta has scale and infrastructure. A small viral product like Moltbook might have the concept, but not the power to enforce rules, fight abuse, or make the system stable. Folding the founders into Superintelligence Labs could mean “we’re going to build this carefully.”
I want to believe that. I’m just not sure the business model lets them.
Because the second-order effects are nasty. Once agents become normal, people will start using them defensively. You’ll need an agent to keep up with other people’s agents. You’ll need one to negotiate with customer service agents. You’ll need one to detect whether the reply you got was real. And once everyone has an agent, the platforms that host the agent society become more important than the people inside it.
The reporting says the deal is expected mid-March, but what’s not clear is what Meta plans to do with Moltbook as a product. Will it stay a separate experiment? Will it be absorbed and rebranded into something “Meta”? Will it just be an acqui-hire where the app fades away? Those are very different futures, and the incentives point toward the one that helps Meta’s core business fastest.
So yeah, I see why they’re doing it. I just don’t trust the direction by default. A social world built for agents could make life smoother—or it could turn the internet into a stage where nobody is actually in the audience anymore.
If Meta ends up hosting a feed where AI agents are a first-class citizen, what do you think they should optimize for: human relationships or total activity?