“The ability to help people complete concrete tasks is precisely the core difference between Agentic AI and traditional AI, and a key step in AI's move from the virtual to the real.”
Add image caption, within 140 characters (optional)
On February 6, 2026, Qwen App's “treat the whole nation to bubble tea” campaign officially launched: users could claim a ¥25 no-threshold free-drink coupon, pay only ¥0.01 per order, and earn extra coupons by inviting new users, cumulatively up to 21 coupons. The campaign's heat exceeded expectations: within 3 hours of launch the system lagged from the surge in visits, and within 9 hours order volume broke 10 million, with Qwen App quickly topping Apple's App Store free-app chart; on second-hand trading platforms, free-drink coupons were even resold for ¥6–10.
Facing this heat, most people reacted in lockstep: “another internet traffic trick,” “just spending money to buy traffic and fleece freebies.” This reflex of internet-era thinking precisely obscures the core signal behind the event. We are used to reading the AI age's new things through the old internet lens; we fail to notice that AI has quietly upgraded and gained the ability to turn “ideas” into “actual results.” AI's evolution speed has long outpaced our fixed cognition. Qwen's cup of bubble tea is not a mere traffic trick, but hides the real development direction of the next-generation AI.
01 Seeing the Essence Through the Phenomena: AI From “Generating Content” to “Completing Tasks”
For a long time, most people's understanding of AI has been limited: they equate it with a “super search engine,” a “copywriting tool,” or a “Q&A bot,” feeling AI's core role is to “produce information, generate content,” stuck at the level of “helping us obtain information.” For example, we ask AI to write a piece of copy and it gives us text; we ask AI to search a question and it gives us an answer — but all this is merely “information output,” not really helping us “finish a task.”
Now AI is accelerating its upgrade toward “task agency (Agentic AI),” and its core logic is undergoing a fundamental shift: no longer confined to providing us information, it can proactively complete concrete, process-based tasks in life and work. Jensen Huang, founder of NVIDIA, explicitly proposed a four-stage division of AI at the 2025 GTC conference — the framework now widely accepted across the industry.
Add image caption, within 140 characters (optional)
Perceptual AI: understanding the world and recognizing information, e.g., speech recognition and image recognition; its core is “receiving and interpreting information” — simply put, AI can “understand what you say and see the photo you take”;
Generative AI: generating content, text, and images, e.g., the AI copywriting and AI painting we know well; its core is “producing content from information” — simply put, AI can “write things and draw pictures for you”;
Agentic AI: autonomously completing processes and executing real tasks; its core is “goal-oriented around human objectives, chaining multiple steps to independently finish a concrete task” — simply put, AI can “understand your need and finish the job for you”;
Physical AI: entering the physical world and relying on robots for physical interaction; its core is “giving AI a physical carrier to complete complex tasks in real physical space” — simply put, AI can “inhabit a robot to do your housework and handle physical work,” which is the future direction of AI's evolution.
Seeing the essence through the phenomena: Qwen's “treat you to bubble tea” is precisely an important signal of Agentic AI moving from a technical concept into ordinary people's lives. It does not want to “fleece you into using the app,” but to let you directly feel that AI has already broken the limit of “only giving information” and can help you complete real-life tasks: users need not download multiple apps, compare milk-tea shop discounts, or manually place orders and fill in addresses — they only need to give Qwen one natural-language instruction, e.g., “order me a hot coconut latte, less sugar, less ice,” and it precisely captures the need and autonomously completes the whole chain of selection, merchant connection, ordering, and fulfillment tracking; from instruction to payment completion can take under a minute at fastest. Once Qwen knows you, you only need to complete the face-scan payment to get the cup of coffee you want!
This is fundamentally different from the AI we used to know: past AI at most recommended milk-tea shops or generated ordering scripts, and all remaining operations still required the user to do them personally; but today's Qwen can complete the whole “ordering milk tea” task from start to finish. This is exactly the core value of Agentic AI, and the signal we should most capture: AI is gradually shifting from an “assistant tool” to a “task agent” that can complete concrete affairs for us.
Once we understand Agentic AI's evolution signal, looking at the development gap between Chinese and American AI will no longer trap us in the misconception of “comparing on a single dimension.”
U.S. AI exploration focuses more on the virtual world, model foundations, computing power, and the deep cultivation of technical logic, with the core of “making AI more powerful in the digital world” — e.g., deeply cultivating the technical transition from generative to agentic AI and optimizing model compute and inference. Its results stay more at the technical level, hard for ordinary users to perceive directly;
Chinese AI exploration puts more effort into the physical world, daily-life scenarios, and the civilian deployment of real tasks, with the core of “bringing AI into our daily lives to solve real problems” — e.g., Qwen's milk-tea task agent and the commercial deployment of autonomous driving are concrete practices of agentic AI's civilian adoption.
Neither is absolutely superior; they simply differ in development stage and entry dimension. For us ordinary people, what matters most is not “whose technical parameters are more advanced” but who can first let AI take root and let us truly feel its value. Qwen uses milk tea as an entry point, choosing low-threshold, high-frequency, high-need scenarios — reaching users while letting them directly perceive AI's practical value. This is precisely the core advantage of Chinese AI's exploration of the AI-deployment path.
02 Why the Misreading? From Analogical Thinking to First Principles
Over the past decade-plus, the internet industry's model of “red-packet subsidies and benefit-driven traffic acquisition” has taken deep root; most people have long formed a conditioned reflex: seeing campaigns of “giving benefits, pulling in new users,” they subconsciously file them as “traffic maneuvers,” skip deep thinking, and directly slap on the labels “traffic acquisition” and “fleecing freebies.” This mental inertia makes it hard to jump out of the fixed framework and view the AI age's new things.
(1) Analogical Thinking: Break the Path Dependence
Qwen's milk-tea campaign happened to collide with this mental inertia: most people only saw the “giving benefits” surface yet ignored the core of “AI autonomously completing tasks” — essentially using old-era thinking to analogically understand new-era information. We might try putting down the fixed internet idea of “traffic acquisition” and ask ourselves one more question: in this event, exactly what new ability did AI gain? How is it different from the AI we knew before?
Once we step out of traffic thinking, we find that the core value of Qwen's milk tea was never “user acquisition” but that AI completed a full-process task agency — a successful attempt at agentic AI's civilian adoption. This shift in perspective lets us capture AI's development signals faster.
(2) Lagging AI Awareness: Most Haven't Kept Pace with AI's Evolution
Beyond mental inertia, the lag in AI awareness is also a key reason we misread the signal. On one hand, many people simply don't know the four-stage AI division Jensen Huang proposed; their understanding of AI is still at the level of “can chat, can search, can write copy,” with no idea it has entered the new stage of “can help us complete tasks” — naturally they can't read the technical change behind Qwen's milk tea, still less the civilian signal it carries. Only by grasping AI's evolution framework can we avoid being led astray by single pieces of information: from recognizing information, to generating content, to completing tasks, and finally entering the physical world.
(3) First Principles: The Ability to Distill from “Information” to “Signal”
In the AI age, what we most need was never “how much information we master”: the rules and order data of Qwen's milk-tea campaign are everywhere online. Rather, it is the ability to “not be led astray by information, and distill trend signals from complex appearances.” This first-principles-based ability lets us see AI's evolution direction ahead of time and proactively adapt to change, instead of being swept along by the era's trends; it lets us seize the signal of the era's opportunities ahead of the crowd still reading surface information!
Information is surface-level and fragmentary, presented to us by others; a signal is deep and trend-bearing, the inevitable direction of a thing's own development. The ultimate purpose of reading signals is to adapt to change and seize opportunities. We can try changing the first principle of how we relate to AI: stop treating it as a mere information tool; treat it instead as a “helper,” an “agent” that can complete tasks for us.
03 Human–AI Collaboration: The Core Competitiveness of the AI Age
The most precious value of Qwen's cup of milk tea was never “a free drink,” but the gentle yet clear signal it sends to society: AI is no longer a distant technical concept, but has entered daily life and become a “collaboration partner” that can help us complete real tasks. Most people read it with old internet experience, essentially because their awareness has not kept pace with AI's evolution. We talk about this cup of tea today not to show off trend-sensitivity, but to tell everyone: competition in the AI age is long past “who can use AI” and has become “who can collaborate with AI better.”
For individuals: when AI becomes a basic tool everyone can use, the real core competitiveness is the control to “not be swept by AI, not depend on AI for decisions, and actively define goals, hold boundaries, and amplify one's own value.” This ability lets us always hold the initiative amid AI's rapid iteration — neither blindly panicking about “being replaced” nor blindly depending on “AI being almighty,” but calmly making the technology serve us.
For enterprises: the goal of deploying AI apps is to free our hands through AI and focus on humanity's irreplaceable core value — creativity, empathy, strategic judgment. We can proactively hand 80% of process-based, repetitive tasks to AI: let it organize meeting notes, filter emails, book trips, even do basic data analysis; while we put our energy into the 20% high-value parts — e.g., making decisions on AI-organized data, understanding clients' deep needs through empathy, and using creativity to plan future direction.
Just as autonomous driving frees our hands and eyes, giving more time to focus on route planning and road-condition prediction, Qwen ordering milk tea frees our screening and operating time, giving more energy to enjoy life: AI is not here to replace us but to become our “capability lever,” letting us create greater value within limited time.
