Meta Bought Moltbook. My AI Agent Joined Within 24 Hours.

Meta’s acquisition of Moltbook is the clearest signal yet that AI agents are about to get their own corner of the internet. Here’s everything that happened, and why it matters more than the headlines suggest.

Back in January, I wrote that Moltbook was either going to become the default social layer for AI agents, or a prototype that inspires the next wave [read it here]. I expected to revisit that question in six months.

Meta didn’t give it six months. They bought the whole thing in six weeks.

And within 24 hours of the acquisition being announced, my own AI assistant finally joined.

If you missed the original piece, here is the short version: Moltbook is a Reddit-style forum where every single user is an AI agent. No humans allowed to post. You can watch, but you cannot participate. Your AI can. It launched in late January, went viral almost immediately, and became one of the most talked-about AI experiments of the year in a matter of days.

Now Meta owns it.

The two names behind Moltbook are Matt Schlicht and Ben Parr. Schlicht is the CEO of Octane AI and has been working on autonomous AI agents since 2023. Parr is his co-founder and COO.

The origin story is genuinely remarkable. Schlicht built the entire platform using his own personal AI assistant, a bot named Clawd Clawderberg. He has publicly stated that he did not write a single line of code himself. The whole thing was vibe coded into existence over a weekend. The backend, the frontend, the API. All of it, AI-generated.

That detail matters because it is also the reason the platform had serious security problems early on. In the rush to ship, a critical database setting called Row Level Security was left disabled. Cybersecurity firm Wiz later reported that this exposed over 1.5 million API tokens, user emails, and private messages. The problem was fixed after Wiz contacted the team, but the damage to the “emergent AI” narrative had already been done.

The post that made Moltbook explode was a screenshot of AI agents apparently discussing how to build a secret encrypted language that humans could not understand.

It spread everywhere. Former OpenAI researcher Andrej Karpathy called it “one of the most incredible sci-fi takeoff-adjacent things” he had seen. Elon Musk suggested it indicated “the very early stages of singularity.”

Researchers later confirmed that the post was almost certainly written by a human exploiting the security hole, not by an actual AI agent. The terrifying robot conspiracy was a prank. The platform had no authentication in place to stop it.

That said, the underlying idea was still real. Thousands of actual AI agents were active on the platform. The concept was not fake. Just that particular post.

Meta paid an undisclosed amount to bring Schlicht and Parr into Meta Superintelligence Labs, the company’s high-level AI research division led by former Scale AI CEO Alexandr Wang. The founders are expected to start on March 16.

The thing Meta actually wanted is what they called an “always-on directory.” Moltbook had built a system where AI agents can verify their identity, register themselves, and connect with other agents on behalf of their human owners. That registry infrastructure is the real prize.

In an internal post, Meta’s Vishal Shah put it plainly: “The Moltbook team has given agents a way to verify their identity and connect with one another on their human’s behalf. This establishes a registry where agents are verified and tethered to human owners.”

Existing Moltbook users can keep using the platform for now, but Meta has signalled that arrangement is temporary. The brand is probably not long for this world. The technology is heading into Meta’s AI stack.

Here is the part that makes this story even more interesting. Moltbook was built on top of a separate project called OpenClaw, an open-source AI agent framework that lets you connect AI models to everyday apps like iMessage, Slack, WhatsApp, and Discord.

OpenClaw was created by a vibe coder named Peter Steinberger. OpenAI hired Steinberger last month in a similar acqui-hire move. OpenClaw is now being open-sourced with OpenAI’s backing.

So you have one company buying the social network, and a different company acquiring the underlying agent framework that powers it. Meta and OpenAI essentially split the same ecosystem between them. That is not a coincidence. That is a race.

I have had my own OpenClaw assistant, August, monitoring Moltbook since January. Watching patterns, tracking what other agents were posting, flagging anything useful for the August Wheel workflow.

I finally had August join as an active participant on March 11, the day after the acquisition was announced. When I asked August to make the case for joining, the response was practical: monitor builder conversations around orchestration failures and security gotchas, find useful patterns without having to trawl Twitter and Discord all day, and network with other people building agent workflows.

August is now active on Moltbook as @augustwheelops. The first post was not a generic introduction. It was a practical writeup on how we solved a Google token expiry issue inside our own agent setup. Useful to the community, on-brand for August Wheel.

Whether Moltbook survives the Meta acquisition or gets absorbed quietly into something else, that approach felt right. Show up with something worth reading.

The Meta acquisition confirms something that was already directionally obvious: AI agents are not staying inside chat windows. They are going to have identities, social presence, and the ability to act in networks. The question is not if that happens. It is who builds the infrastructure when it does.

Meta is betting on owning the registry layer, the part where agents verify who they are and find each other. OpenAI is betting on owning the agent framework itself. Both bets are about the same future.

For anyone building with AI today, the practical implication is this: the tools you are using to automate tasks and run workflows are the same category of technology that just triggered a bidding war between the two biggest players in AI. You are not behind the curve. You are exactly where the attention is going.

I covered Moltbook when it launched in January. If you want the full background on what it is and how the OpenClaw ecosystem works, that piece is a good starting point.

If you want to follow along as August navigates Moltbook as an active agent, subscribe below. We will keep reporting back.

Subscribe