Using individual efforts to secure China's ticket to the AI competition.

Text: Whale Brother, Whale Select AI

The speed at which Manus has broken through the circle has set a new record, surpassing those set by ChatGPT and DeepSeek, going viral across the internet within just one day of its internal testing!

For this company based in Wuhan, named "Butterfly Effect," this is actually their second blockbuster product. The first was the AI product Monica, which has 10 million users overseas and an annual recurring revenue (ARR) reaching tens of millions of dollars, making it one of the most revenue-generating AI products of 2024.

Manus is the team's first AI Agent product, and during its internal testing on March 6, it was praised by many industry professionals as a product that "opens the era of Agents" due to its autonomous task execution capabilities. From its explosive popularity to today, it has only been four days, and even without public testing, the product has already reached a high internal testing code price of 50,000 yuan based on a few showcased cases.

In fact, this company was offered $30 million by ByteDance for acquisition in 2024. However, they ultimately felt the price was not fair, and the acquisition did not go through. Whale Brother has learned from multiple sources that the reason was that ByteDance considered the Monica product to be somewhat "shell-like," with high customer acquisition costs and poor retention data, fearing it might be replaced by larger models in the future, while they themselves would launch a Doubao plugin, hence the lower offer.

For Monica, an ARR of tens of millions of dollars with a 3x PS (using the concept of market sales rate) acquisition is several million higher than their previous valuation, which clearly does not meet the expectations of the investors behind it.

In the era of mobile internet, the company and its products were sold to a certain unicorn. After the AI era began, from Monica to Manus, the new company's product strength and execution capability are quite formidable, leading to support from investors like Sequoia China and Tencent for their further entrepreneurship. After all, in a landscape where major companies lack blockbuster AI-native products, they have created two breakout products with minimal funding.

Especially with Manus's breakout, accompanied by astonishing praise, many believe it represents innovation at the level of DeepSeek, serving as the key to opening the door to AI Agents and even a representative product on the path to AGI; at the same time, it has also faced a tide of ridicule, with many claiming it lacks core technology and is merely a shell product; others suspect it is a marketing scheme involving KOLs, as it hasn't made much of a splash abroad; and some believe the subsequent cost of opening services will be too high, making it difficult to implement.

Amidst the mixed opinions, Whale Brother has also conducted in-depth research on this product, attempting to communicate its breakthroughs and significance from a more objective and long-term perspective.

DeepSeek breaks the computing power paradigm, Manus opens the era of Agents

On March 6, at the small launch event for Manus, the company proclaimed "The next ChatGPT moment," indicating that the door to the Agent era has been opened.

Whether Manus is the next GPT moment is still uncertain. DeepSeek, which has been popular for a month, has already proven its strength.

DeepSeek exploded in popularity during the Spring Festival because it was the first time people experienced the charm of reasoning models. It is indeed very intelligent, with many answers demonstrating depth, breadth, and flexibility that exceeded previous products like Doubao and Yuanbao. This is the advantage of reasoning models over directive models, and it showcases the wisdom of DeepSeek as a company.

In 2025, large models will face three breakthrough directions: large parameters, multimodal, and reasoning. The first path, Grok 3, is still robustly breaking through the scaling law of large models with 200,000 H100 GPUs, while others are strengthening through architectures like MoE; multimodal is the main focus for many, with representatives like OpenAI abroad and Doubao, Tencent's Hunyuan, and Jieti Xingchen domestically, integrating DIT architecture, reportedly excelling in listening, seeing, and writing; the third path is enhancing model intelligence through reinforcement learning (RL) on the reasoning side, with DeepSeek representing this direction, and Tongyi quickly following suit.

DeepSeek's larger characteristic is its outstanding cost control ability, especially with the subsequent five days of open-source projects. This gradually confirms the capability of optimizing underlying infrastructure, breaking the technological financial order established by companies like Nvidia based on "computing power."

What is "computing power paradigm"? The financial hegemony of the United States has undergone three fundamental logical transitions: from gold, to oil, to computing power, essentially reshaping the credit system through the monopoly of global core resources, maintaining the hegemony of the dollar.

Gold standard collapse (1944-1971)

The Bretton Woods system established a link between the dollar and gold, but the insufficient gold reserves of the U.S. led to the collapse of the system. In 1971, Nixon announced the decoupling of the dollar from gold, necessitating a new anchor for the dollar.

Petrodollar hegemony (1974-present)

The U.S. secretly negotiated with Saudi Arabia to settle oil transactions in dollars, establishing a "petroleum-dollar-U.S. debt" cycle: oil-exporting countries earn dollars and then purchase U.S. debt, making the dollar the global reserve currency. At its peak, 86% of global oil trade was settled in dollars, and the Federal Reserve harvested global wealth through the dollar tide.

Rise of the computing power paradigm (2020s-)

In the digital age, computing power has become a new means of production. Nvidia's H100 chip has become "computing power currency," with the global computing power market reaching $2.6 trillion in 2023, and U.S. companies holding 60% of the market share, with computing power replacing oil as the new pillar of dollar credit.

The astonishing demand for computing power from large models has contributed to Nvidia's stock price increasing by over 435% in the past two years, with its market value rising from $300 billion to nearly $3 trillion in the past decade.

However, for Manus, it has not yet caused a stir in the foreign tech circle. It cannot influence Nvidia's stock price fluctuations like DeepSeek. But according to partner Zhang Tao, the company has only over 50 employees and managed to create this viral AI product in just two to three months.

WeChat search index comparison: Manus has not surpassed the peak period of DeepSeek

But in the domestic market, Manus is still the hottest AI product; just by releasing a few examples on its website, it has sparked a storm of discussion online. In the face of high demand, many people are even seeking codes at high prices. An application for generating Manus invitation codes even topped the iOS paid chart in the Chinese region on March 8. Of course, this product is not particularly useful, but it has capitalized on the traffic.

Manus's insight leads, major companies lack innovation

For most people who have not experienced OpenAI Deep Research (which costs up to $200 per month), the domestic Manus is indeed somewhat stunning upon first use. Whale Brother tested the following question:

For a question like creating an embodied intelligence report, the first attempt got stuck while analyzing the data, and the final PPT request could not be fulfilled, so the second attempt requested a text version report, which also stopped before creating charts.

Currently, it seems that Manus often cannot understand and control its own capabilities, frequently biting off more than it can chew.

From a theoretical standpoint, Manus is not complex. Manus integrates computer use, virtual machines, and multi-agent collaboration into an AI product.

From Baoyu AI

But the most important breakthrough of Manus is that it has achieved the productization of general domain Agents compared to automation programming software like Devin and bolt.new.

Employees who previously worked at this company mentioned on social media that the company has a deep accumulation of engineering practice and agent workflow:

In September-October 2023, they first launched an agent in China, and the todolist.md from that time reflects the best practices learned from various agent solutions.

In March 2024, they developed the GPTs platform, and since early 2024, they have been accumulating technology related to browsers, gaining a deep understanding of browser context utilization.

Starting in November 2023, they began working on search, accumulating capabilities for agents to connect to the internet for information. I was not involved in that part.

In July 2024, they gained experience in increasing social traffic through roast.

In November 2024, they developed an understanding of coding capabilities across various models in their coding product.

"Indeed, each aspect is relatively thin, but the combination of these building blocks formed during this window of opportunity is also strong," this employee commented.

Whale Brother believes that Manus's greatest success lies in its product insight that surpasses major companies.

Xiao Hong, the founder of Butterfly Effect, as a serial entrepreneur, previously launched the "Yiban Plugin," a WeChat plugin product with millions in revenue (the Whale Select account's new media operation is also using it), and then seized the opportunity in corporate WeChat SCRM to create the Weiban Assistant. With the rise of AI, he developed the large model aggregation product Monica, and now with the surge of AI Agents, he has launched Manus.

Especially the success of two consecutive AI-native products is not easy. Even strong players like OpenAI have not been as successful with their other products outside of ChatGPT, many of which remain in a semi-finished state, such as GPTs, SearchGPT, DALL.E, Whisper, etc.

Currently, domestic major companies' AI products generally lack creativity. From AI social to AI search and AI coding products, they are all products that others have as well.

This image was generated by Tencent Yuanbao AI

Meanwhile, the Butterfly Effect company, after seizing the first wave of AI plugin dividends with Monica, has also taken the lead in creating the first mature product in the Agent space. In the five stages of AI shared by Sam Altman, L1 (chatbots), L2 (reasoners), L3 (agents), Manus has successfully positioned itself as an early breakout product in the L3 stage.

Their philosophy of "Less Structure, More Intelligence" has led them to abandon the path of AI browsers that cannot compete with major companies, allowing them to discover new opportunities.

As for whether this wave of breakout effect is generated by marketing leverage, Whale Brother does not believe so.

Previously, when Monica released its Chinese version, there was a wave of collaboration with KOLs, distributing free membership usage quotas through KOLs. When Manus invited KOLs to participate in a small launch event for discussion, they indeed described it as "the world's first general-purpose Agent product," but I learned that there were no cooperative promotions with KOLs.

It is estimated that if Manus had not made a big splash, it would have released free token quotas through KOLs like Monica. Now it seems that is no longer necessary, but from the previous WeChat search index, it clearly hasn't reached the level of DeepSeek.

Wu Bingjian, a partner at Soul Capital, once described: After DeepSeek broke through, society was indeed educated by the facts, and everyone began to compete in technical levels—how to improve Attention, how to enhance MoE, how to mix FP8 and FP16 training. Further, it became a competition in originality—who can innovate the next generation of model architectures, who can discover the next set of training methods.

From this perspective, Manus has also pushed everyone’s product strategies back to the forefront of innovation, focusing on what innovative products can be created after the AI application/Agent era in 2025, rather than just a battle for traffic in AI assistants.

Shelling does not affect, but success is still far away

At this stage, AI products essentially lack core barriers and competitiveness. Product concepts are difficult to register as patents, and engineering capabilities are where major companies excel. Therefore, Whale Brother previously mentioned on Xiaohongshu to guess which major company would create a similar product.

However, to be fair, compared to some previous general Agent products, such as OpenAI's Operator, Anthropic's Claude, and Tencent's APPAgent, Manus is a more refined Agent product in terms of engineering delivery.

However, refinement is not a high barrier for a product. After Manus became popular, the MetaGPT team spent three hours developing OpenManus and open-sourced it.

After Manus explored actual needs, optimized technical engineering paths, and continuously improved product functionality details, it is indeed not difficult for outsiders to "half-open" and copy the homework. If it can be replicated in half a day, does that mean Manus has no barriers at all?

An investor lost the opportunity to invest in Perplexity due to the shelling theory.

Currently, countless AI search products have not affected the development of Perplexity. The native product understanding will help Perplexity continuously update better features, while other similar products can only follow up and replicate later.

The same goes for Manus. At this stage, what it most needs to solve is to accept financing from a major company. After all, the reason for not publicly testing is that the server capacity is insufficient.

Messages released during a media communication meeting indicated that the Manus team also provided the cost of single-task operation: about two dollars. The cost has dropped to 1/10 of DeepResearch, but the single-task cost is still nearly 15 yuan. This is also the reason why Manus has adopted a small-scale invitation code distribution, which has caused internal system crashes.

Accepting financing from major companies is not just about funding; more importantly, it is about the low-cost supply of large model APIs. According to media reports from the communication meeting, Manus primarily calls the Claude model, along with some fine-tuning of open-source large models. Will it accept investments from Alibaba, Tencent, or ByteDance in the future, similar to Kimi's funding + resources investment?

This would allow Manus to launch a subscription service with an annual fee of less than a thousand yuan. Otherwise, if the price exceeds this, it may become a toy for a small number of professionals.

On the other hand, improving product details and service capabilities, and rapidly iterating is fundamental to ensuring vitality.

Currently, generating a response with Manus still takes a long time, and many tasks still crash. Manus needs to put more thought into productization, rather than what Manus product manager Zhang Tao said: "It's really very simple; there are no secrets; it's just believing in the power of the model."

The model is the foundation, and product details are the service capability. For example, Claude 3.7 Sonnet has once again broken the ceiling in coding capabilities, but Cursor, with its advantages in code auto-completion, can still attract users to pay for subscriptions.

More importantly, the MCP (Multi-Agent Communication Protocol) aggregation model is already showing high growth potential. This is also the development path that Manus should adopt in the future.

From the speed of large model evolution, built-in agents may also be a trend. If GPT-5 can achieve a fusion of reasoning and directive models, multimodal capabilities, and built-in agents, it may become unexpectedly powerful. Major domestic companies should also be following this route; before that, Manus needs to scale up its user base and revenue.

Summary

DeepSeek has broken the narrative that only foreign models succeed in the large model field and has demonstrated extremely low cost implementation capabilities, making everyone believe that the mysterious power of the East has the ability to impact the global technological financial order. In other words, DeepSeek has single-handedly secured China's ticket to the AI competition, directly ending the era of global investors slashing valuations on Chinese assets.

For Manus, it showcases the strongest form of AI native, not a cookie-cutter Chatbot, nor an AutoAgent that resembles a Trojan horse, but rather something more useful across various scenarios, with hope for implementation.

To put it grandly, it may become the WeChat of the next AI era, but it may not be able to influence the popularity of foreign products like Facebook. To put it modestly, it has opened the public's awareness of Agents and given many startup teams the confidence to continue dreaming.

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。