A version of this story appeared in CNN Business’ Nightcap newsletter. To get it in your inbox, sign up for free here.

Meta, the company that changed its entire brand identity five years ago on the promise of a technology it turned out no one wanted, just spent an undisclosed sum to acquire Moltbook, a “social network” built for “AI agents.”

I’m quoting Moltbook there, but I’m also using quotes because it’s important to remember that artificial intelligence is just lines of code and therefore not capable of socializing. AI “agents” are just bots that can mimic human actions online, so perhaps the better descriptor of Moltbook is a “pseudo-performance of a social network”? If using Facebook is an online performance, Moltbook is an attempt at aggregating every social media user into a kind of meta-spectacle of socialization.

Whatever you want to call it, Moltbook is essentially a Reddit-like forum where bots are meant to “talk” to one another. And Meta’s decision to buy it is significant — not so much because of what it Moltbook is but because of what it represents: a speculative bet on a novel technology with no demonstrated utility for humans.

The project went viral last month for two major reasons:

For those inclined to see the AI race as a bubble, Meta’s decision to buy an unpolished Reddit imitator on the expectation — still speculative at this point — that “agents” are the future of digital ads and commerce is a clear sign of the kind of irrational exuberance that has defined all of history’s dumbest market meltdowns, from tulip mania to the dot-com crash.

Meta last year shelled out $2 billion to acquire Manus, an AI developer, and $14.3 billion to acquire data-focused startup Scale AI, and it has reportedly dangled $100-million-plus compensation packages for top-tier researchers. And like its Big Tech rivals, Meta is borrowing tens of billions a year to finance these deals — another possible sign of overconfidence.

For Meta, the bet on Moltbook lies partly in hiring its creators, Matt Schlicht and Ben Parr, who are joining the Meta Superintelligence Labs unit. Perhaps the pair can expand Moltbook into something that drives revenue, or perhaps they create another viral hit.

Certainly, they’ve proven they can generate buzz, and in the world of AI, where valuations are lofty and almost fully detached from business fundamentals, that hype matters. (Meta reportedly tried to recruit Peter Steinberger, the creator of OpenClaw, but rival OpenAI got to him first, per Business Insider.)

“The Moltbook team joining MSL opens up new ways for AI agents to work for people and businesses,” a Meta spokesperson told CNN. “Their approach to connecting agents through an always-on directory is a novel step in a rapidly developing space, and we look forward to working together to bring innovative, secure agentic experiences to everyone.”

Bottom line: Meta is all in on AI, much the way it was all in, five years ago, on its conception of an online town square called the Metaverse. Moltbook may not be a repeat of that flop — Meta yanked its Metaverse unit’s funding and slashed its staff earlier this year — but its aggressive AI spending should be a reminder of the company’s history of getting out over its skis in the name of a shiny new object.

CNN’s Hadas Gold contributed reporting.