Meta's Muse and Qwen Struggle for Adoption Without Real Demand

According to a report by Huxiu’s Fang Wei Kan Shi Nian, Meta shares rebounded from a low after the unveiling of its personal AI agent Muse, bringing its market capitalization to about $2 trillion.

The piece says there is no doubt that Meta itself holds intrinsic value worth $2 trillion.

However, it argues that expecting Muse to achieve mass adoption like Instagram Reels or Threads, with its 500 million monthly active users, and to generate cash through commercialization is unrealistic.

It warns that disappointment is likely a year from now, and that investments in upstream computing resources and NPU-related plays will meet the same fate.

The essay positions personal AI agents represented by Muse and Alibaba’s Qwen app as products not born from genuine demand for a use case.

It describes them as strategy-led launches driven by big companies’ fear of being left behind, mixed with an intent to attract capital-market attention.

It stresses that management’s strategic needs differ from the pressing needs of ordinary users.

It states that piggybacking investments and development will ultimately lead to losses.

Skepticism Toward Personal Agents

Muse and the Qwen app promise to “get errands done.”

Food delivery, ride-hailing, hotel booking, and price comparison are cited as examples.

But in everyday life in China, food delivery and ride-hailing have already reached a high level of maturity through mobile networks and dedicated apps.

The essay describes this state as already optimized to a score of 90 out of 100.

It argues users have little reason to pay for efficiency gains from 90 to 95 points.

Ordinary users’ behavior after work is not oriented toward efficiency.

It centers on spending time staring at smartphones, the piece says.

Few are calling for an always-on AI that runs 24 hours a day, 365 days a year, remembers months of preferences, and makes suggestions on its own.

Only a minority of the public feels the need to have AI remember long-term personal goals, it says.

Fragmented needs such as search, shopping, and video viewing account for most demand, it says.

The Public’s Real Demand Lies in Free

Services and Entertainment

The essay argues that the pressing demand behind widely adopted products can each be explained in a single sentence.

For WeChat, the decisive factor was being able to send text and voice messages and make calls for free.

For Baidu Search, the core was being able to look up information for free and reach world events in one second.

TikTok won support because users can watch highly immersive videos without thinking.

All of them were characterized as free, high-frequency, and strongly compelling.

By contrast, there is no clear explanation of whose problem and which problem Muse solves, the piece says.

Examples such as haggling over discounts with a purchasing agent are unlikely to hold up, it says.

If merchants identify an inquiry as coming from an AI agent, they will block it, leaving no room for negotiation, it says.

Demand that cannot be explained is not real demand, it concludes.

For 1.3 billion of China’s 1.4 billion people, there are few errands in daily life that need handling, it says.

A Product Shaped as a Strategy-First Experiment

Meta’s and Alibaba’s moves are described as strategy-driven, not user-driven.

The piece quotes Meta’s chief executive himself acknowledging that people do not know what they can have AI do.

The remark is interpreted as the flip side of providers also not knowing the right answer.

If they knew, they could win simply by spreading the solution directly, it says.

The current situation is an experiment of throwing out solutions to see which use cases remain, it says.

The biggest obstacle to people’s use of AI is not inadequate AI capabilities, but that people do not know what they can have AI do.

As a use case for Muse, planning to bake sweets with a three-year-old daughter was introduced.

It involved gathering ingredients with tools such as Instacart and having the AI plan steps suited to a toddler and a beginner adult.

Another proposal introduced was having it create a strategy guide for the game “Civilization” for the daughter, expanding into learning the history of different civilizations.

Both were described as limited use cases for a sophisticated segment.

They were judged as highbrow, narrow in scope, and distant from the daily lives of the masses.

According to base materials, the Qwen app connects directly with Taobao instant shopping and the main Taobao app, Amap ride-hailing, and Fliggy travel.

Alipay can be used for payments.

Meta’s payments will require connections with third parties.

Qwen is seen as having an advantage in the depth of integration.

Even so, after investing tens of billions of yuan, the “get errands done in one sentence” use case has yet to take off, it says.

Three Areas Where AI Works Today

The essay narrows effective applications in the AI revolution at present to three.

They are conversational responses, programming, and office work.

They have taken hold one by one each year, it says.

Information retrieval, programming, and office processing directly substitute for intellectual labor, it says.

The work of lawyers, doctors, teachers, engineers, programmers, and designers will be impacted, it says.

There is little point in pitting AI against AI in existing businesses that have already reached 90 points, it says.

Competing on efficiency in food delivery or ride-hailing has a low probability of success, it says.

Agreement was expressed with longtime Silicon Valley observer Ben Thompson’s formulation.

Silicon Valley has to relearn two things every 10 years. First, consumers do not pay for process. Second, consumers do not care about productivity.

Dropbox was cited as a typical example.

It was excellent as a consumer product, but after its business stalled, it was rebuilt from the ground up and pivoted to the enterprise market.

Consumers do not pay, but companies pay because employees’ time has economic value, it says.

Consumers after work do not want to become even more efficient, it says.

Remarks by TikTok’s former head on an audio program were also quoted.

Asked whether TikTok will exist in the AI era, he said he does not know.

To overturn TikTok, a new product vessel is needed, he said.

It cannot happen within the vessel of full-screen immersive short video, he said.

As long as AI creations are compatible with TikTok, creators will post to TikTok, which is superior in traffic and revenue, he said.

The view is that without innovation in the vessel, no replacement will occur.

Household Machines and the Long-Term Outlook

The essay views the present as an extension of the first through third industrial revolutions.

Smartphones, computers, electric lights, and electric shavers in use today are extensions of the electrical revolution.

The massive storage, computation, and long-distance distribution of information are attributed to the power of semiconductors and the information revolution.

Mainframes, minicomputers, personal computers, networks, smartphones, short video, and cloud computing are described as applications in the same lineage.

The AI revolution is an enhanced version of the information revolution, producing near-intelligent processing through large-scale, complex computation, it says.

The “story machine” will become one setting where AI intelligence lands, it says.

Conventional story machines could only tell stories fed into them and could not be called intelligent.

AI-driven story machines will become smarter and could replace parents and grandparents in the storytelling role, it says.

Only the sense of companionship cannot be replaced, it says.

Household machines fit the ultimate vision of application, it says.

Expectations for household machines go beyond laundry and cooking, it says.

That is because skills covered by washing machines and rice cookers costing a few hundred yuan have already been replaced.

They will be expected to look after children, serve as conversation partners after work, bring in laundry when it rains, call for emergency help and assist transport when elderly people fall, and even provide emotional value, it says.

Replacing domestic workers earning nearly 10,000 yuan a month will be the benchmark, it says.

The ultimate use case is still far off, it says.

Editorial Opinion

On short-term impact, excessive expectations for personal AI agents will be corrected in the next three to six months. Sympathy buying of upstream components based on announcements by Meta and Alibaba will lose momentum. Product managers will face demands to polish user flows that are already at 90 points, such as integration with payments, maps, and commerce. If disclosure of usage metrics remains sparse, capital markets will take a harsher view. Development resources will return to areas that generate payment, such as conversational responses and office processing. On the long-term view, over one to three years, products with a use case explainable in one sentence, like free messaging, search, and entertainment, will retain the advantage. AI agents will be absorbed into enterprise office work and programming assistance, and their independent form as consumer resident assistants will shrink. Applications with a concrete vessel, such as household machines and story machines, will become the next focus. Efficiency competition without a vessel will be weeded out. Design rooted in users’ life context will determine value. Questions from the editors are noted. It is asked whether true demand is limited to what can be explained in one sentence. It is asked whether the 90-point wall, where efficiency is said to have no value, can be broken through integrated payment-linked integration.

References

• “ 没有杀手级场景的产品是很难大规模流行的 “, by 方伟看十年© — 虎嗅网, 2026-09-27T23:13:01.000Z (ARR)
• Source URL: https://www.huxiu.com/article/4894159.html?f=rss