Silicon Valley's still flexing with massive model sizes and billion-dollar funding rounds, while Chinese tech giants are straight-up battling over... 1-cent bubble tea....
Just wrapped Chinese New Year 2026, and yeah—AI era officially hit China, but not in that shiny sci-fi way. It's more like "Hey Qwen, hook me up with a milk tea for basically free."
Alibaba's Qwen dropped 3 billion RMB (~$400M+) on this insane promo: update the app, get a 25 RMB no-threshold voucher, say one sentence like "帮我点杯奶茶" and boom—AI handles the whole thing across 300k+ stores (Mixue, Luckin, Nayuki, etc.).
Pay literally 1 cent. Orders exploded—10 million+ in the first 9 hours, servers crashed, bubble tea shops got swamped (riders lining up, some stores had to pause orders lol). People were posting screenshots everywhere, it was chaos in the best way.
Tencent's Yuanbao threw 1B in cash red envelopes, Baidu Wenxin did 500M, ByteDance's Doubao crashed the Spring Gala with AI integrations and gifts.
Total burn this holiday Like 80-100B rmb+ across the board.
In China, if your AI doesn't sink into real daily life (ordering food, booking tickets, shopping), you're dead.
No one's paying for fancy poetry-writing bots. Alibaba's playing both sides: open-sourcing Qwen 3.5-Plus (top open model globally) while grinding hard on everyday stuff—one prompt and it closes the loop on tickets/milk tea/groceries.
Byte's betting big on video gen with stuff like Seedance 2.0 scaring actual video teams—they're gambling that soon we'll just consume endless AI-generated content streams instead of chatting.
Harsh truth: Chinese AI was born with a profit gun to its head. These giants are taking on massive compute debt just to force the habit shift. If your AI isn't the default "go-to" on phones, you're the next Nokia.
A lot of people clown on China's "practical" AI approach, but I think it's the real killer move. Remember 10 years ago? We basically taught the whole country mobile payments... by spamming red envelopes.
Real AI wins aren't about who writes the best poem—it's who can handle your sudden milk tea craving and route 130 million orders without breaking a sweat.
So, hot take: Should AI be out there "changing the world" first... or just reliably get your damn bubble tea order right?
@thsottiaux I've recently noticed that Codex suddenly shuts down and ask me to log in again. Why is this happening? Has anyone else experienced the same issue?
*The Art of War* states: There is no instance of a nation benefiting from prolonged warfare. When applied to the current competitive landscape of the AI market, both Anthropic and OpenAI face a persistent threat from open-source models; the longer this continues, the greater the likelihood of their defeat.
I have been reflecting recently on the relationship between keywords and actual user needs—specifically, the genuine core needs.
Users often start by using the most general term that broadly covers what they are looking for, typically opting for a well-known brand; this is simply human nature.
However, if they are dissatisfied or encounter issues after using the product for a while, they will search again—this time with a more specific intent, which gives rise to "long-tail keywords."
At the same time, factors such as education level, communication style, region, and language result in a wide variety of ways to express the same underlying intent.
Consequently, you might find that different keyword phrases—despite having different wordings or structures—reflect the same intent but vary in search volume based on collective user behavior.
Yet, Google’s semantic recognition capabilities allow it to clearly understand that these phrases all represent the same need.
Therefore, if you build a website targeting a keyword with lower search volume, you face less competition and can achieve rapid ranking growth; you may then find that the site also ranks well for a high-volume keyword that expresses the same intent.
This reveals that the concept of finding "blue ocean" demand within a "red ocean" market goes beyond simply identifying a niche market scenario. Within the Google ecosystem, you can also carve out a successful niche in a saturated market by targeting keywords that express the same intent through different phrasing.
It feels like the stream of new technologies surrounding AI agents has dried up recently. In the early days—with MCP, tool calling, skills, and loop engineering—new technical concepts seemed to emerge every few months; now, however, progress appears to have slowed, and major companies have fallen back into the familiar cycle of focusing on building increasingly powerful models.
@paulg It is a massive strategic blunder; for a startup, spending a few hundred dollars on a domain—and allocating the remaining $249,900 to community building and targeted Google ads—would yield far greater value than sinking the entire budget into the domain name.
Here’s a funny thing: plenty of people on X are hyping up Kimi, DeepSeek, and GLM, yet the vast majority of AI coding users within China actually rely on Codex and Claude.
I've noticed a pattern on X: your profile must feature impressive life experiences, jobs, status, or achievements; otherwise, no one will pay attention to you. But this is normal, people always want to focus on people of value.
There are plenty of cafes, restaurants, and bars here that are great for gatherings. They're not hard to find(download https://t.co/BI1P22B1hX, You can find the contact information for these merchants on Dianping. You can see it as another version of Google Maps), but you'll need to confirm the specific date and number of people in advance. I am sure you will find amazing people living in Shanghai. I would love to help you with answering all questions although I am living in Chengdu.