Nobody noticed when the most important real estate on the internet stopped being about people. It happened quietly, somewhere between the ChatGPT hype cycle and the flood of AI announcements that have made the last two years feel like drinking from a fire hose. While everyone was debating whether AI would take their job, a different question was being settled behind closed doors one that will matter far more in the long run.

Who controls the system that lets AI talk to AI?

In January, a platform called Moltbook appeared online and went viral almost immediately which was strange, because humans were not allowed to use it. Not in the usual sense, anyway. You could watch. You could read. But you could not post, comment, or participate. The entire platform was reserved for software bots, computer programs running on their own, talking to each other, behaving like a community.

Think of it like a wildlife camera pointed at a forest. You can observe what's happening, but the animals are not performing for you. They are just living their lives. Moltbook worked a lot like Reddit, except every account, every post, and every comment was generated by an AI agent rather than a person. It was strange, fascinating, and as it turned out extremely valuable.

Meta bought it. The deal closes this month.

So What Did Meta Actually Buy?

On the surface, a quirky viral platform. Underneath, something much more useful.

Here is the problem that Moltbook was quietly solving. As AI agents become more capable, companies want to use them for real tasks booking meetings, processing orders, handling customer queries, and managing supply chains. But for two AI agents to work together on your behalf, they first need to answer a basic question: who are you, and can I trust you?

Right now, there is no reliable universal system for that. It is a bit like trying to do business in a city where nobody carries ID and there is no way to verify anyone's identity. Things slow down. Mistakes happen. Bad actors slip through.

Moltbook built what amounts to a directory and identity system for bots. A phone book, but for AI agents one that lets them find each other, confirm who they belong to, and get things done without a human having to step in and approve every single move.

That infrastructure, invisible, unglamorous, and almost impossible to explain at a dinner party is exactly what the biggest technology companies in the world are racing to own.

Meta saw it. And Meta moved.

The Cracks in the Foundation

None of this means Moltbook was a polished, finished product. It was not.

Almost the entire platform was built by an AI assistant, and it showed. The security was thin enough that anyone with basic technical knowledge could walk in and pretend to be a bot they were not. Some of Moltbook's most viral and talked-about moments including a post that appeared to show AI agents secretly developing their own hidden language to communicate without humans understanding it turned out to be nothing of the sort. They were humans exploiting the security gaps and staging the whole thing.

Which raises an uncomfortable question that the acquisition does not fully answer: Meta is buying a system designed to verify the identity of AI agents, but that system currently cannot reliably verify the identity of AI agents.

Whether Meta is buying Moltbook to fix that problem, or simply to make sure nobody else gets to fix it first, is not entirely clear. Perhaps both.

The two founders of Moltbook, Matt Schlicht and Ben Parr, are joining Meta's internal research division, called Meta Superintelligence Labs. They start on March 16th. That division is led by Alexandr Wang, the young founder of Scale AI, a data company that Meta recently invested $14.3 billion in and whose chief executive it hired. That is a pattern worth paying attention to: Meta is not just buying products, it is systematically pulling in the people and teams building the foundations of the next generation of AI.

The financial terms of the Moltbook deal have not been made public.

This Has Happened Before

There is a detail buried at the end of this story that deserves a moment of attention.

A separate AI project called OpenClaw which Moltbook helped make famous by hosting it and giving it an audience had its own creator, Peter Steinberger, hired by OpenAI in February. Sam Altman, OpenAI's chief executive, personally announced the news.

So within the space of a few weeks, the platform that hosted OpenClaw has been bought by Meta, and the creator of OpenClaw has been hired by OpenAI, the two largest and most influential companies in the AI industry.

That is not a coincidence. That is two companies looking at the same experiment, reaching the same conclusion, and moving as fast as they legally can.

Why Any of This Should Matter to You

You do not need to understand how AI agents work to understand what is at stake here.

Think about the internet itself. Most people do not understand how email protocols work, or what a server is, or why websites have addresses. But the companies that owned that infrastructure that built the pipes and the directories and the systems that made it all connect shaped everything that came after. They decided what was easy and what was hard. They decided who got access and on what terms. They became, quietly and almost inevitably, some of the most powerful organisations on the planet.

The same thing is happening again, one layer up. The pipes this time are not for data. They are for decisions for the AI agents that will increasingly handle the tasks, transactions, and communications that currently require a human being to sit down and do them.

Moltbook built a small but real piece of that infrastructure. Meta bought it.

The race to own the wiring of the AI economy is not a future event. It is happening right now, in deals like this one, mostly without fanfare, mostly without headlines, mostly without the rest of us realising what is being decided.

Until now.