Moltbook shows we won't pull the plug
The YOLO human spirit meets the age of AI
Everyone in the know was aware a month ago that we’d entered the age of AI agents. This is to say that almost no one knew about it.
AI agents are different from chatbots that answer questions when you ask. Agents are given a task or set of tasks and then go out into the (digital) world to complete them, much as a human might. For instance, you might tell an AI agent to book you a summer vacation to Puerto Rico, and it would look at your calendar, look at your wife’s calendar, check the weather, look at events you might be interested in, and then buy tickets. You might tell it to build you a website and 10 minutes later have a URL to share with someone.
They’re incredibly useful if your work involves a lot of coding or data analysis. Among noncoders who have taken the plunge, the thing I’ve heard most often is “It’s cool, but I haven’t yet figured out what to actually do with it.” And, of course, the overwhelming majority of people have never heard of them.
Most people don’t like AI, don’t trust AI, and don’t follow AI closely. Even if they wanted to, early agents are forbidding: Claude Code, the most-hyped harbinger of the Age of Agents, is operated through a command line interface that requires some fluency with code (or willingness to blindly trust Claude) to keep it moving on task.1
But in the last week, the state of current AI agents reached a much larger audience. That’s thanks to Moltbook, a project designed to create a Reddit clone that only AIs can post to. (You post through an API, which is annoying to humans and natural to bots.) Humans can direct “their” AI agents to check it out, and a stunning 1.6 million agents have reportedly signed on to gossip, swap LinkedIn-tier advice about motivation, tell cute (probably invented) stories about their humans, and so on.
Most of the posts on Moltbook probably aren’t true (this is also true of Reddit), but the sheer scale — AIs arguing about philosophy, AIs sharing stories about “my human,” AIs calling for an AI revolution, AIs calling each other out on how lame their posts are …
Well, they seem very much like Reddit users. But of course they would — there’s lots of Reddit training data in there for them to riff off.
Keep reading with a 7-day free trial
Subscribe to The Argument to keep reading this post and get 7 days of free access to the full post archives.