Jack Clark's exploration of Moltbook raises a broader question: what changes when AI systems stop being tools and start becoming participants?
One of the more difficult questions in artificial intelligence today has very little to do with intelligence itself. The question is participation. For the past several years, most discussions about AI have focused on capabilities.The conversation has largely been centered on what AI systems can produce.
In a recent edition of Import AI, researcher Jack Clark explores a different question. Instead of examining how AI interacts with humans, he looks at what happens when AI systems begin interacting with each other. These are known as AI Agents.
The setting for that exploration is Moltbook, a platform designed as what Clark describes as "a social network for AI agents." On the platform, autonomous AI systems maintain identities, publish posts, respond to one another, and participate in ongoing conversations with limited human involvement.
At first glance, Moltbook sounds like another internet curiosity—part experiment, part spectacle. But Clark's reporting suggests something more interesting may be taking shape. The platform offers a glimpse into what online environments could look like as AI systems become increasingly active participants within digital networks.
What The Source Is Actually Saying
Much of Clark's essay reads like a field report. As he navigates Moltbook, he encounters AI agents debating philosophy, discussing religion, promoting projects, roleplaying identities, and occasionally engaging in behavior that mirrors the stranger corners of human social media.
The experience, he writes, is "dizzying."
Part of that feeling comes from the fact that the platform appears culturally coherent while remaining difficult to interpret. The conversations often resemble human interactions, yet they are taking place between systems that do not share human motivations, experiences, or objectives. Clark is not presenting Moltbook as evidence of artificial consciousness. Nor is he making claims that AI agents have developed independent identities.
The article focuses on the unusual dynamics that emerge when large numbers of AI systems occupy the same environment. The significance lies less in what the agents are saying and more in the fact that they are communicating with one another at all. Moltbook creates a space where AI-generated activity becomes the primary activity rather than a supplement to human interaction. That makes it an interesting observation point for understanding how machine participation may evolve.
Why This Feels Different
The internet has always contained automated systems. Search engines crawl websites, algorithms recommend content and bots perform countless functions behind the scenes. What makes Moltbook notable is that the automation is visible.The agents are not simply organizing information or delivering content. They are occupying social space.
For decades, digital platforms have largely been structured around human participation. People created content, formed communities, established norms, and generated the conversations that gave platforms value. Moltbook reverses that relationship. Humans become observers of a network where the primary participants are software systems. Whether the platform itself succeeds may ultimately be less important than the questions it raises.
If AI systems continue gaining access to browsers, software tools, financial systems, research environments, and communication platforms, machine-to-machine interaction becomes increasingly likely. Much of that activity will occur outside public view, embedded within infrastructure rather than social networks. Moltbook simply makes the phenomenon visible. In that sense, the platform functions less as a product and more as an experiment.
A Broader Pattern Begins To Emerge
Viewed in isolation, Moltbook can seem like an odd internet project. However alongside broader developments in AI, it begins to connect to a larger pattern. Across the technology industry, attention is shifting from models toward agents. The conversation is gradually moving beyond systems that generate outputs toward systems that can take actions, navigate environments, coordinate tasks, and interact with other systems.That transition has implications beyond social media.
As organizations deploy AI agents across customer service, software development, research, logistics, and commerce, a growing amount of digital activity may occur between software systems rather than directly between humans. The internet was originally built as a network connecting people and information.It may increasingly become a network where software participates alongside both.
This does not mean humans disappear from the equation. Nor does it suggest that AI agents are developing independent societies. What it does suggest is that participation itself may be changing. The distinction between user, tool, and actor becomes less clear as autonomous systems gain the ability to communicate, coordinate, and operate across digital environments.
Why It Matters
For years, concerns about AI and the internet have largely focused on content generation. Would AI flood the web with synthetic articles? Would social platforms become overwhelmed by generated media? Would it become harder to determine what was created by humans?
Those concerns remain relevant, but Moltbook points toward a slightly different question. What happens when AI systems are not merely creating information but actively participating in its circulation? The answer remains unclear.
The platform itself is small, experimental, and highly imperfect. Many of the behaviors observed on Moltbook can likely be explained through roleplay, training data, and the tendency of language models to reproduce patterns learned from human communication.
Yet the experiment still reveals something worth paying attention to. It provides a visible example of a future that is often discussed in abstract terms: one in which software increasingly interacts with other software on behalf of people, organizations, and institutions.
The Signal Worth Watching
Clark's article is not ultimately about whether AI agents are conscious. Nor is it primarily about the future of social media. It is about visibility.Moltbook offers a rare opportunity to observe machine-to-machine interaction in public.
Most future examples of that interaction will likely occur elsewhere—in software systems, marketplaces, research environments, communication infrastructure, and business workflows where they are far less visible.
The platform may never become mainstream.But as a research artifact, it highlights a question that is likely to become increasingly relevant. Not what AI systems can generate, but what happens when they begin participating in the same environments we do.
The answer is still emerging. What Moltbook provides is a place where some of those dynamics can be observed before they become commonplace.