Text | Kaori

Editor | Sleepy.txt

On December 30, 2025, Meta acquired Manus for over $2 billion.

Three months later, just last night, it quietly bought Moltbook. This time, the price was not announced.

What is Moltbook? On January 28, 2026, developer Matt Schlicht launched a strange website that looks like Reddit, but the only rule is: only AI Agents can post, humans can only observe.

In the two weeks after its launch, Moltbook briefly became a talking point in the AI circle, a space where humans could only watch while AI conversed amongst themselves, satisfying Silicon Valley's impulse for a post-human social experience.

But the hype came quickly and left just as fast. In the subsequent six weeks, the AI universe was immersed in a flurry of new hotspots and crayfish crazes, and the daily active user data for Moltbook had already fallen back to baseline; the AI Agents in the community continued to post, but there were hardly any human viewers left. It was at this moment, almost forgotten, that Meta bought it.

This is Meta's third major AI acquisition in the past year. The world's largest social company is burning money at a rate of $100 billion a year but is increasingly struggling to answer a fundamental question: what does it really want to become?

Moreover, more and more onlookers feel that Zuckerberg always seems to be behind the times. But this judgment, in fact, reverses the issue.

When Everyone on the List Says No

Zuckerberg is not missing the mark, nor is he not offering a high enough bid. The reality is that those he truly wants are no longer in need of him.

Beginning in the spring of 2025, Zuckerberg reportedly initiated a personal recruitment campaign that can only be described as unprecedented. He met with candidates at his private residences in Lake Tahoe and Palo Alto, offering signing bonuses of up to $100 million.

Those contact points included AI search engine Perplexity AI, the then most important independent company in AI video generation, Runway, Ilya Sutskever’s new company Safe Superintelligence after leaving OpenAI, and OpenAI's former CTO Mira Murati's new startup, Thinking Machines Lab.

All four of them declined.

This list of refusals speaks more about Meta's situation than any successful acquisition could.

Founders in 2012 and 2014 faced an arithmetic problem: what could they grow to while remaining independent? With Facebook's distribution, how many users could they jump straight to? The answers were almost obvious, so Systrom and Koum sold.

That was an era when distribution was still scarce, and Meta happened to have the world's largest distribution channel.

Founders in 2025 face a different problem. Sutskever left OpenAI to build a company based on his judgments about AI safety, a judgment he was not prepared to give up within any organizational structure. Murati, similarly, founded Thinking Machines.

Aravind Srinivas from Perplexity left OpenAI, Google Brain, and DeepMind, starting his venture in 2022; he does not need Meta's distribution, he needs independence.

In the eyes of this group in the AI era, capital is no longer a scarce commodity. The independence of the narrative is what truly matters.

After being simultaneously rejected by these four companies, what did Meta gain?

Scale AI. A data labeling company that has never trained a large model independently. Its core business is organizing human labelers to classify and tag data. This is the infrastructure for AI training, a shovel-selling business, but it is not AI research itself.

This $14.3 billion deal, nominally a strategic investment, is essentially a shell game, moving Scale AI’s 28-year-old founder, Alexandr Wang, into Meta.

At the same time, in the specific track of the Agent ecosystem, OpenAI made a similar move but secured a different person.

OpenClaw, the underlying framework for Moltbook, was an open-source AI Agent tool set up by Austrian developer Peter Steinberger in an hour, allowing users to run their own AI Agents locally and control them through applications like WhatsApp and Telegram. After launch, OpenClaw's GitHub stars surpassed 200,000 within weeks, with weekly visits reaching 2 million.

Moltbook was born from the OpenClaw ecosystem.

In February 2026, OpenAI hired Steinberger. Sam Altman called him a genius on X and announced he would lead the company's next generation of personal Agents. OpenClaw entered an independent open-source foundation supported by OpenAI.

Steinberger later revealed that Meta had also approached him, and Microsoft had as well. In the end, he chose OpenAI, with the only condition being that OpenClaw must remain open source.

In the same Agent ecosystem, OpenAI took the engineer who built the framework while Meta acquired the one who built the platform using that framework.

What Did Buying People Buy?

Before Wang arrived at Meta, there was one person who had been there for twelve years.

Yann LeCun, a Frenchman, Turing Award winner, and one of the deep learning triumvirate alongside Hinton and Bengio. Recruited by Facebook in 2013, he founded FAIR, transforming an ad-driven social company into a credible AI research institution in academia.

He has a judgment he repeatedly states publicly: large language models are a dead-end. The future of AI lies in world models that can understand the physical world, capable of perception, memory, reasoning, and planning, rather than engines predicting the next word on a massive corpus of text. He is not performing dissent; each public statement reiterates this one belief, never vague.

In June 2025, Alexandr Wang arrived. Meta announced it would acquire 49% of Scale AI for $14.3 billion, and Wang would become Chief AI Officer, leading the newly established Meta Superintelligence Lab. Meanwhile, LeCun was asked to report to Wang.

There is a basic fact that needs to be clarified: Wang's Scale AI has never trained a complete large model. Its core competitive edge is high-quality data labeling, providing training data for models like GPT, Gemini, and Claude. This is an indispensable part of the AI industry chain, but it is a different matter from training models themselves.

LeCun did not accept this reporting relationship. In November 2025, he announced his departure to establish a new company, AMI, to continue research on world models. Meta stated it would cooperate with AMI.

This outcome can be interpreted as a normal management change. But it also signifies something more precise: Meta's bet on the direction of LLM has become irreversible and no longer accommodates the internally most qualified dissenting voice. A Turing Award winner who believes the current route is wrong and a 28-year-old founder executing this route cannot coexist in the same reporting chain. Zuckerberg made a choice, opting for the latter.

How did it turn out?

Of the initial 14 researchers in Llama, 11 have already left Meta. Internal dissatisfaction due to bureaucratization and direction confusion led to about 600 layoffs in October 2025, which Wang described as correcting previous bureaucratic expansion.

According to the Financial Times, Wang expressed to those around him that Zuckerberg's micromanagement was stifling, and their relationship became tense. Scale AI's original clients, Google, Microsoft, and xAI, began to withdraw, fearing damage to neutrality; Scale AI's acting CEO had to publicly issue a letter emphasizing the company's independence.

The strategic partner Meta spent $14.3 billion acquiring immediately turned into a credibility-damaged collaborator after the purchase.

One more thing. Llama 4 Behemoth, Meta's most important flagship model, has already completed training. However, the internal evaluations did not meet expectations, the release plan was shelved, and discussions about whether to open-source it are ongoing.

An organization with expected capital expenditures exceeding $100 billion annually cannot roll out its flagship product on time.

At this time, what did Meta do? It bought Manus, and then bought Moltbook.

Meta Was Once the Most Prolific Spender

In April 2012, Instagram had just released its Android version. On the launch day, the server crashed due to a surge in traffic. The next day, Zuckerberg made a phone call, offering $1 billion.

At that time, Instagram had only 13 employees and 30 million monthly active users. It had been launched for only 18 months.

This deal was considered impulsive by many, and Zuckerberg himself later made a remark that became ironic: "We do not plan to make many acquisitions like this, if there are any."

We all know what happened later. Ten years later, Instagram's monthly active users exceeded 2 billion, making it one of Meta's most profitable assets.

WhatsApp is the second version of the same story. At the time of acquisition in 2014, WhatsApp had 4.5 million monthly active users, more than Twitter at the time.

With 55 employees, it handled 50 billion messages daily, with daily active users accounting for 72% of monthly active users, while the industry average was 10% to 20%. Facebook personally wrote in the official announcement: "WhatsApp's message volume is approaching the total volume of global telecommunication SMS." This was a statement of a fact that had already occurred.

Sequoia Capital gained around 5000% return from this investment, and the media at that time described Facebook as having made a huge investment.

Both transactions share a common structure. Before the acquisition, the target product had already completed the hardest part: self-validation.

Instagram proved that mobile photo sharing was an irreversible user habit, and WhatsApp proved that instant messaging could replace the entire telecommunication SMS system. What Meta did was use its distribution channels with a billion users to push established entities to another level.

Facebook, which had not yet changed its fate, was not a wave creator; it was the one who ran up the fastest after the wave arrived.

Snapchat was the only miss in this logic. In 2013, Zuckerberg offered $3 billion, but Evan Spiegel declined. But Meta spent two years copying the Stories feature to Instagram and WhatsApp. Since then, Snapchat has had no more room for growth.

If you can’t buy it, copy it. If you can’t copy it, hunt it down. This strategy was invincible during that era.

The problem is that that era has ended.

Meta Has No Dreams, 2026 Edition

In 2018, tech journalist Pan Luan wrote an article titled "Tencent Has No Dreams," with the core conclusion being that Tencent replaced its intent to create products with investments and acquisitions. This article was widely circulated within Tencent.

That article was written eight years ago. It was about Tencent, but the symptoms have not been eradicated within Tencent.

Tencent later found its way out, not by acquiring more companies. WeChat grew from within, a product pried out by Zhang Xiaolong in the cracks of a large organization, redefining Tencent's era position.

Where is Meta's WeChat?

Meta AI's monthly active users reached 1 billion in early 2025. This number sounds impressive, but monthly active users do not equal definition.

ChatGPT changed people's understanding of the term AI assistant in 2022, altering search habits for 100 million users within two months. Gemini embedded itself into Google Search and the Android ecosystem, allowing almost all Android users to interact with it unknowingly. Anthropic's Claude became the top choice for enterprise AI deployment, boasting a clearly established trust advantage in finance and healthcare.

What is Meta AI? It is a feature residing in Instagram and WhatsApp. There are 1 billion people who have used it occasionally, but no one has changed anything because of it. No one has rethought their work methods or redefined the boundaries of AI because of Meta AI.

The situation with Manus is somewhat subtle but also worth examining. The company's selling point is its ability to autonomously perform complex multi-step tasks like market research, resume screening, and stock analysis with a general Agent. It sounds more substantive than Meta AI, but Manus's Agent capabilities run on Anthropic's Claude.

Meta spent $2 to $3 billion to acquire a working AI Agent, while the core intelligence of this Agent comes from one of its competitors. At the level of foundational model capability, Meta has yet to position itself to be self-sufficient.

Now, looking back at Moltbook, its true role becomes clear.

Matt Schlicht, who dropped out of high school, came to Silicon Valley, interned at Ustream, and later co-founded Octane AI with Ben Parr, an AI marketing tool aimed at e-commerce brands, focusing on personalized recommendations and customer interaction automation for Shopify sellers.

This is a business with commercial logic; both of them are active voices in the AI Agent community: Parr is an AI columnist for The Information, and they co-host AI courses, co-manage an early-stage AI fund called Theory Forge, investing in startups like Gumloop and Wordware in the Agent ecosystem.

They have real connections and influence in this community, which is what Meta truly wants to acquire; Moltbook itself is merely an addition.

But the problem is, they are not Peter Steinberger.

Steinberger built an OpenClaw prototype in an hour; this framework reached 140,000 GitHub stars in two weeks, becoming one of the most crucial underlying infrastructures in the Agent ecosystem. He was recruited by OpenAI because he had specific technical ideas and building capabilities.

Schlicht and Parr's positions in the Agent ecosystem are storytellers and connectors, not builders. This distinction is not meant to belittle; it is merely a harsh reality. Meta and OpenAI secured different kinds of assets in this talent competition.

This gap is a passive result, not an active choice. Perplexity declined, Runway declined, SSI declined, Thinking Machines declined, and Steinberger chose OpenAI. What remains willing to come is from those who are willing to come.

The asset pool that Meta can reach now is no longer on the same level as in 2012.

That year, Zuckerberg offered $1 billion, and after some consideration, Instagram's founders felt that leveraging Facebook's distribution was the fastest path to a leap forward, so they signed. That question had a unique rational answer.

Today, AI entrepreneurs enjoy independent narratives, have ample capital, and a clear understanding of what they can do; they understand what selling to Meta means.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。