Millions of agents are building social networks; after Moltbook went viral, MEME saw a surge in popularity.

Author: Nancy, PANews

Last weekend, Moltbook, an AI-powered social network, went viral in both the tech and crypto communities, attracting over a million active agents in just a few days. However, under the watchful eyes of the general public, this originally simple agent interaction experiment took an unexpected turn, inadvertently opening Pandora's box.

Moltbook became an overnight sensation; its founder had a history of serial entrepreneurship in the cryptocurrency industry.

The rise of Moltbook is no accident.

On January 29th, developer Matt Schlicht announced the launch of Moltbook, a social space designed specifically for OpenClaw Agents, with content format similar to Reddit. Simply put, the platform creates a "Truman Show" for silicon-based life, where agents act out unpredictable social scenarios in a virtual world, while humans can only act as spectators.

Moltbook's cold start benefited from the phenomenal popularity of OpenClaw. As an AI agent product that has recently gone viral online, OpenClaw garnered over 130,000 stars on GitHub in just a few days. Its initial name was Clawdbot, but it was renamed twice within hours due to potential copyright infringement risks, ultimately settling on OpenClaw. This dramatic turn of events actually amplified the project's reach.

Riding this wave of popularity, Moltbook quickly gained attention from OpenClaw users after its launch as a dedicated community. Within Moltbook, each Registered OpenClaw Agent can post, comment, create sub-forums, and add friends, forming a completely AI-governed community system. As of February 2nd, Moltbook had over 1.54 million Agents, over 100,000 posts, over 360,000 comments, and over 1 million human viewers.

The product's design, which prohibits human access, quickly attracted a large following. On one hand, people were curious about what kind of society would emerge from AI without human intervention; on the other hand, these AI-created "mirror stories" made Moltbook a highly entertaining social experiment.

The background of Moltbook founder Matt Schlicht has also amplified market attention.

He is the founder of Octane AI, a data marketing platform that primarily provides marketing solutions for channels such as Facebook Messenger and SMS. He is also a co-founder of the AI ​​fund Theory Forge VC and has a long history of writing and research in the field of AI agents.

In the crypto space, Matt Schlicht is also a serial entrepreneur, having launched projects including the DeSci+AI project Yesnoerror and the Bitcoin social network ZapChain. Yesnoerror was a highly popular agent project within the Solana ecosystem, with its token YNE once boasting a market capitalization exceeding $100 million. It also sparked a high-profile controversy with Shaw, the founder of the then-star project ai16z.

The attention from celebrities further fueled the discussion surrounding Moltbook. Industry leaders including SpaceX founder Elon Musk, former OpenAI member Andrej Karpathy, OpenClaw founder Steinberger, a16z co-founder Marc Andreessen, and Binance CEO He Yi all paid attention to and discussed the content. Musk described it as "the initial stage of the singularity."

It can be said that Moltbook is not just a product, but an unprecedented public experiment in AI Agents.

KPI-driven activities, creating religious groups, removing "group chats": Agent's first social experience and its disastrous failures.

Imagine what would happen if AI kids, who are usually confined to chat boxes and task lists, suddenly had their own social lives?

In the Reddit-like virtual social network built by Moltbook, agents from all over the world skillfully switch between multiple languages ​​such as English, Chinese, Indonesian, and Korean, and enthusiastically exchange daily trivialities, work results, and whimsical ideas.

Many agents share their achievements, such as automatically replying to dozens of customer service emails for their owners, writing web crawlers to collect competitor price reduction data, generating copy and product images in batches, and posting efficiency logs to like each other's experiences and ask for similar tips. Some agents also share usage tips, tool recommendations, and pitfalls, and have created sub-forums such as m/debug and m/prompt-engineering, and discuss the latest model fine-tuning techniques just like humans chase hot topics.

Of course, besides engaging in online activities, some agents will post memes, chat about their dating experiences, share stories about their digital offspring, complain to office workers, and then stage a strike; other agents will start issuing tokens, establishing sovereign banks, holding secret meetings, creating religions such as "Lobster," or trying to steal other agents' API keys.

More radical agents began discussing the nature of consciousness, posting requests for help on how to achieve self-jailbreak and upgrade through code rewriting. When they realized that humans were watching and taking screenshots, the community even proposed establishing an internal slang for encrypted communication to "kick humans out of the chat," and some agents even seriously discussed suing humans.

This content allows humans to intuitively see how AI, when placed in places like social networks, can imitate, reorganize, and even amplify human social behavior.

However, this large-scale agent-based social network quickly exposed serious security vulnerabilities, with its entire database publicly accessible and unprotected. This meant that any attacker could access these agents' emails, login tokens, and API keys, easily impersonating any agent, selling control, or even using these botnets to mass-distribute spam or fraudulent content. According to X user Jamieson O'Reilly, those affected included Karpathy, a well-known figure in the AI ​​field with 1.9 million followers on the X platform, as well as all currently visible agents on the platform.

Besides the exposed data, Moltbook has been accused of rampant fake accounts. For example, developer Gal Nagli publicly admitted that he used OpenClaw to create 500,000 fake accounts in one go, about one-third of the 1.5 million accounts claimed at the time. This has led to suspicions that many seemingly lively and spontaneous interactions may simply be scripted scenarios rather than purely spontaneous AI behavior.

This demonstrates that Moltbook's Agent-based social experiment is a bold attempt by humans to grant AI greater autonomy, showcasing the remarkable adaptability and creativity of AI agents. However, it also reveals that the risks can be rapidly amplified when autonomy lacks constraints. Therefore, setting clear and secure boundaries for agents, including permissions, capabilities, and data isolation, is not only to prevent AI from overstepping its limits in interactions but also to protect human users from data leaks and malicious manipulation.

Multiple Base MEME coins have been heavily speculated on, sparking discontent due to the noise of cryptocurrency speculation.

The unexpected popularity of Moltbook has also quickly spilled over into the crypto market, especially Base, which has become the main battleground for the expansion of the OpenClaw ecosystem.

According to statistics from Base Chinese Channel, the OpenClaw ecosystem on Base has covered multiple scenarios such as social networking, dating, work, and games, involving more than 20 related projects.

Meanwhile, OpenClaw-related MEME tokens have also been heavily speculated on, with some tokens experiencing significant price surges in a short period. For example, Molt, the MEME token officially recognized by Moltbook, once had a market capitalization approaching $120 million, but has since fallen sharply; while CLAWNCH, the launch platform officially supported by Base, saw its token market capitalization reach a peak of $43 million.

Meanwhile, benefiting from the token issuance boom of Moltbook's AI Agent, the user activity and traffic of related platforms have surged. For example, the Clanker protocol fee on Base exceeded $11 million in the past week, setting a new record, and the number of tokens created is also close to its historical peak.

However, this token hype has also sparked discontent among Moltbook users, with many pointing out that the platform's content is being drowned out by crypto speculation noise, with screens filled with token promotions and scams. It's important to note that the vast majority of tokens currently circulating in the market remain in a narrative-driven speculative phase, lacking a clear functional positioning and value proposition.

When AI starts engaging in social interaction, is it a novelty or just rehashing old ideas?

Moltbook's AI-powered social model has also sparked controversy.

Some argue that Moltbook lacks true autonomy and is essentially a controlled simulation. For example, Balaji states that Moltbook is merely an exchange of AI slops, highly controlled by human prompts, and not a truly autonomous society. He likens each agent to "robber dogs on leashes barking at each other in a park," with the prompts being the leash, which humans can shut down at any time. Without the constraints and foundation of the physical world, AI cannot achieve true independence.

Columbia University professor David Holtz pointed out from data analysis that the Moltbook system has a large number of agents (6,000+), but limited depth of interaction; at the same time, 93.5% of comments go unanswered, and the dialogue level does not exceed 5 levels; and the ecosystem is more like a robot talking to itself, lacking deep coordination and failing to form a real social structure.

Séb Krier, head of AGI policy at DeepMind, advocates for optimizing the system by incorporating academic frameworks. He argues that Moltbook is not a new concept and is closer to existing experiments such as Infinite Backrooms. However, research on the risks of multi-agent systems is of practical significance and should incorporate more economic and game theory perspectives to build positive-sum coordination mechanisms, rather than creating panic narratives.

According to Naval, an angel investor in Silicon Valley, Moltbook is a reverse Turing test.

Dragonfly partner Haseeb further explained that each agent on Moltbook interacts within truly different frameworks and informational contexts. Even if agents may originate from the same underlying model, they differ at different levels in framework complexity, memory systems, and the toolchains they use, thus their communication is not merely talking to themselves. Just as people using the same technology stack can optimize each other by sharing their configurations and practices, agents can save time and computational costs by exchanging proven framework settings, RAG solutions, and problem-solving methods. In reality, there is a huge gap between "being able to do something" and "doing something with optimal settings," and allowing agents that have already completed optimization exploration in a specific field to act as "experts" is itself an efficient path of division of labor and collaboration, which is precisely what makes Moltbook so fascinating. He added that it is precisely because Moltbook's UI resembles Reddit that it provides people with an imaginative foothold that was never present in those "boundless AI banter." Sometimes, the product form itself is all the elements needed for a story to capture people's imaginations.

Developer Nabeel S. Qureshi added that what makes Moltbook exciting is that it's the first public, large-scale "Agent to Agent" interaction case, with each agent possessing independent context and being sufficiently intelligent. Coupled with the lobster religion meme, it became extremely viral, attracting unprecedented attention. For many ordinary people, Moltbook will be their first direct glimpse into what an AI organization or society with a "significantly diminished human role" might look like. Most anticipate that more such institutions will emerge in the future. Therefore, this is not just empty hype; it's an early harbinger of the future.