TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Manus Season 2, The Prodigal Returns

Manus Season 2, The Prodigal Returns

2026.08.12
Share
A startup starting anew and a product with an ARR exceeding $300 million.

Authors: Gao Honghao, Song Wei

A startup setting out anew and a product with ARR exceeding $300 million.

On the evening of August 11, the sentence "Manus is now part of Meta," which had hung on the Manus website for nearly half a year, changed to "Manus is about to resume independent operations."

From acquisition to confirmed split, less than eight months passed in between.

Figure: Manus official website before and after August 11, 2026.

In July 2025, Xiao Hong left Wuhan, where he had lived for a long time, and moved to Singapore with the entire team. On December 29 of the same year, he decided to sell the company to Meta for over $2 billion and serve as a Meta Vice President.

This was the largest acquisition in the large model application field at the time, with the deal taking only about 10 days from contact to signing; it was also the third-largest merger and acquisition in Meta's history, second only to WhatsApp and ScaleAI.

One week after the transaction was announced, a spokesperson for China's Ministry of Commerce said at a regular press conference that they would work with relevant departments to conduct an assessment and investigation into the consistency of this acquisition with relevant laws and regulations such as export controls, technology import and export, and foreign investment.

The dramatic turn of events sparked enthusiasm for spectating; far more people participated in the discussion on social networks than had ever used Manus. "Opened the champagne too early" and "$2 billion vanished overnight" were repeated. Some also viewed it as a turning point—the end of founding startups in China and exiting via American capital.

During this period, venture capital institutions that used to squeeze in through any crack to fight for equity cautiously kept their distance. Xiao Hong, Ji Yichao, and Zhang Tao made daily video calls with the team remaining in Singapore to synchronize progress and arrange work.

Amidst the turmoil, the Manus product maintained 1-2 feature updates per week, comparable to the peak at the end of 2025 and faster than a year ago. However, there were no major version number changes; the product stayed at 1.6, focusing more on feature improvements, adding connectors (integrating global products like GMail, Shopify, Slack, etc.), and launching clients. A Manus insider said the development pace remained normal, version 2.0 was already developed, and would be released after the transaction was fully concluded.

On April 27, the National Development and Reform Commission issued a decision: "A decision to prohibit investment was made regarding the foreign acquisition of the Manus project, requiring the parties to revoke the acquisition transaction." Meta subsequently responded in a statement sent to media that the transaction fully complied with applicable laws and looked forward to the review being properly resolved. Transaction revocation (deal unwind)—usually refers to revoking a completed transaction and restoring the pre-transaction state as much as possible.

A few days later, the actual split between Meta and Manus began. According to media reports, both sides stopped data sharing, and Manus employees could no longer access Meta's internal systems.

By June and July, investment institutions that had kept their distance began to realize this was a profitable opportunity. One investor told us that if there was still a funding gap, they would very much like to participate.

Some astute VCs and entrepreneurs began to act; they prepared funds, hoping to participate in the first round of financing after Manus returned.

Because during the three months when the turmoil occurred, Manus revenue grew by about 5 times, rising from $300,000 per day to nearly $1.5 million, once surpassing DeepSeek.

During this time, various rumors circulated continuously. In July, reports emerged that Tencent would take a controlling stake in Manus. Thus, the Manus Beijing and Singapore teams maintained daily communication to soothe emotions and stabilize the team. During several months of turmoil, not a single person left this company of 150 people.

People in Beijing had a new rhythm of life. Xiao Hong relearned mathematics and researched large models. Zhang Tao learned rock climbing and practiced Rubik's Cube, spending half a year to reduce his weight from 75 kg to 60 kg, sleeping enough 7 hours every day, even his blood pressure went down.

These few months redrew the trajectory of this startup and also let its founders think clearly about what they really wanted.

Back in December 2025, Manus was raising funds, and Meta sent people saying they wanted to meet the core team to see if there were business cooperation opportunities. After several rounds of communication, during a meeting attended by all core team members, unexpectedly, Meta expressed its intention to acquire in front of everyone. By this time, multiple partners and core technical personnel had followed Xiao Hong for many years and experienced two startups. Selling the company was a way to account for and reward everyone.

After the acquisition concluded, on New Year's Day 2026, Xiao Hong walked by the seaside in Singapore, feeling that the tortuous ten-year journey had reached its end, and next he would go to work at a big company. But when truly arriving at this seemingly perfect endpoint, he felt somewhat lost in his heart.

And after the acquisition was clearly going to be revoked, the huge reversal and blow in the eyes of the outside world instead gave these serial entrepreneurs renewed expectation: their startup story had entered Season 2.

According to someone close to Manus, Xiao Hong stated during a team conference call that he did not want to spectate or recall this era, but wanted to continue placing himself in the wave and participating in it. Otherwise, he would just become a person writing an autobiography recalling those years.

Manus is a classic tech startup story. Before ChatGPT, Xiao Hong used the earlier released GPT-3 large language model, saw the possibilities of the technology, pulled Ji Yichao and Zhang Tao into the partnership, and made a browser shell for GPT called "Monica".

It is hard for a browser to let people see the possibility of the next super product. One entrepreneur met one of Xiao Hong's partners a year before Manus was released, talked about the already released Monica, and felt the other party was quite unconfident.

During the development of Monica, along with the progress of large models, this team found a forward-looking product form through trial and error, letting the agent work fully automatically like a human. By 2024, Cursor, Copilot, and even ChatGPT were still assuming people staring at screens, constantly sending prompts to direct AI work. At that time, model inference costs were an order of magnitude higher than now, but Manus was already doing Wide Research (broad research) consuming massive tokens, betting that tokens would become particularly cheap. They were also one of the earliest AI application companies to build their own Evaluation (evaluation) team, establishing a loop of closed-loop development.

These forward-looking judgments and investments kept Manus leading and growing for a long time. In early 2025, shortly after Manus was released, a major tech company internally initiated at least 7 similar projects, but did not take away its users. During the investigation period, Manus ARR broke through $300 million for the first time.

Xiao Hong once used "Trade-Industry-Technology" to summarize his understood startup path: first find the market, then build product and business, finally walk towards the depth of technology. Through Manus now he has already completed the first two steps—making a product needed by a large number of users, and also establishing a real business.

Manus's opponents today all grasp the most core technology. In the past half year, Claude and Codex competed fiercely, from model to application, rapidly perfecting using agent to control computers to complete tasks. The desktop application and Plan Mode of planning first then executing launched by Manus during this period were both catching up with these two Silicon Valley pioneers. Recently, a group of domestic large companies also regard work on computers as the next battlefield.

After the transaction is completed, Manus will also face a choice: whether to continue striving for leadership with product insights, or invest in more core technology, such as model training, to meet the cruellest challenge of large model applications.

On the evening of August 11, Manus officially confirmed it would resume independent operations. The Meta acquisition has entered the substantive revocation phase, Manus is about to resume independence; but the final funding repurchase amount, whether the company will return to the Chinese capital system, have not been officially announced yet.

That night, Xiao Hong wrote a letter to the team and investors: "This is my third trip." They mentally view the new Manus as a startup setting out anew: the starting point is a team of over 150 people, a group of partners who remained united through the turmoil, and a product with ARR exceeding $300 million. The story of Manus is far from over.