I am very glad to meet CZ in Tokyo. This is his return to Japan after 7 years, and during this time, the entire industry has undergone tremendous changes.
At this event, I asked CZ a question: How do you view the on-chain of everything and the tokenization of stocks?
He replied: There have not yet been large-scale coin-stock products on the market. The possible reason is that each country and region has different policies, making it difficult for coin-stock investors to receive dividends and easily redeem stocks. There are also price differences during trading. However, projects on the chain are usually global in nature, which was a significant pain point back then.
Restrict areas or try to solve it?
But he reiterated that the trend of RWA cannot be stopped. $BNB
An 8-year veteran practitioner tells you: How to spend the industry's garbage time
The market has gradually become quiet recently. Many colleagues have expressed that they can’t see the direction clearly or are forced to be friends with time. This is actually not the first time. Looking back, since I entered the circle in 2016, I seem to have experienced four major garbage cycles. Although I often comfort myself in my heart that all this will pass, the positive cycle will come back, and liquidity will come back, but when we are really in the garbage time, we will feel more or less depressed or helpless. Winter in Zhongguancun. The first garbage time was when I entered the circle. At that time, the industry had just shifted from the cold winter of 2015 to the recovery period of 2016. OKCoin and Huobi, located in Zhongguancun, also stopped layoffs. The market was gradually recovering, and exchanges finally had positive income.
Multiple articles published last night and this morning about DeepSeek boss Liang Wenfeng’s internal remarks to investors were taken down throughout the afternoon.
According to the full text of the remarks that the media obtained and leaked, DeepSeek had previously asked all attendees to prohibit recording audio/video, screen captures, and sharing the content externally.
An era comes to an end. BitMEX, once the largest crypto derivatives exchange in the world, has just suddenly announced that it will cease operations and shut down on September 23. The company was co-founded by legendary trader Arthur Hayes, known as “lao hei” (Old Black), and others.
In February of last year, BitMEX was reported to be seeking a sale, but no final deal was reached. BitMEX is known for launching perpetual contracts in 2016.
Driverless taxis are back—and can now operate on domestic roads again. Bloomberg reports that China has reissued operational permits for autonomous taxis. This year in March, Baidu’s 100-plus Robotaxis operating in Wuhan were suddenly halted by a computer crash, which led to traffic congestion and prompted regulators to urgently suspend them across the industry.
However, after completing safety reviews by the end of June, some cities gradually resumed approvals and allowed autonomous vehicles back on the road.
That’s so terrible. The guy who previously posted about his 3 million RMB annual salary at the Tencent (Goose) company was later let go, and they just released an internal notice.
I didn’t expect the product mid-level managers in the WeChat department to have such high pay. Back then, many people said it was AI-generated, but it turned out to be real.
A source close to Lianxian Magazine has claimed that within the current Trump administration, there is disagreement over how to respond to China’s AI. Some officials at the White House advocate tightening restrictions and studying how to prevent open-source activity that would encroach on U.S. companies’ market share.
However, the Ministry of Commerce believes these restriction measures are difficult to implement and that it would be better to promote leading U.S. AI labs to develop their own open-source large language models, competing with China in the open-source space.
In the span of a single night, the VC community and peer circles were all forwarding Founder of DeepSeek Liang Wenfeng’s four-hour investor meeting remarks. It feels like a pilgrimage. I’ve organized some of the statements that everyone quoted the most.
• Humans can keep learning, but for the same thing, you have to give AI all the context—which is almost impossible—so AI cannot really replace employees.
• For a normal person, what they need isn’t a computer, but manpower.
• Anthropic is now ahead of OpenAI—not in the long term, but in a phase. In the future, OpenAI and Google will most likely alternate in rising.
• There are too many companies in China that build models. Each one does largely the same thing. In the end, things will certainly converge, but it will take a process.
• We usually don’t work overtime. First, doing research requires a relatively relaxed environment. Second, we’re very focused. We have many products that aren’t fully developed, but we don’t go out of our way to fill in the gaps—that’s also a kind of restraint in our culture.
• Open-sourcing has benefits. This AI thing is big enough that, in the end, it might take up ten percent of human society’s GDP. If we want to monopolize that benefit, we will definitely be left behind by history. This is an objective law—this is a historical perspective.
• Selling APIs isn’t that compelling. You only need a few people to maintain the API. There might not even be customer support—users will come on their own.
• We hope the next generation of models can help our own developers. Put simply: the first goal of the models we build isn’t for everyone to find them easy to use, but for us to use them well. This is the fastest way to achieve AGI.
Is Duan Yongping preparing to swoop in to buy the dip? The so-called “Chinese version of Buffett,” Duan Yongping, said that he has already started trading bearish options on Micron Technology. Next, he plans to sell bearish options on SpaceX, Tesla, and Palantir. He is also considering continuing to add to his holdings of Google and Berkshire Hathaway.
He still likes AI and U.S. tech stocks, but believes that some stocks are currently priced too high. Last time, he also used options to build his position in Pop Mart.
After taking their place within the Ding, the Mìnèi, and the Tuánnèi, a small essay from a big-company employee who works in the Tencent AI team has also arrived inside Bao. An older brother who has been certified as a Tencent in-service employee posted and criticized Tencent Yuánbǎo, accusing that the current frontline middle managers’ requirements for those below are not about professionalism, but about being obedient.
And if someone has objections or opinions regarding middle-management decisions, then they will be seen as a troublemaker to target, and in the end they will be driven out for not meeting output and job-level standards..
Familiar gameplay is back again. AMD announced it will invest up to $5 billion into Claude’s parent company, Anthropic. As part of the deal, starting in early next year, Anthropic will massively purchase AMD AI chips for use in its own data centers.
How does it feel—like the neighbors over at OpenAI, Nvidia, and Oracle have already played this same kind of game?
Didn’t expect that a DeepSeek intern’s daily wage of 5,500 RMB is real. However, because this pay is much higher than that of full-time employees, several senior internal staff members need to score the applicant before an offer can be made.
In normal circumstances, the internship pay is still 500 - 1,000 RMB per day, which is about 11,000 - 22,000 RMB per month.
Nike finally reclaimed its online sales rights. Nike’s online distributor, Top Sports, received a notice that, starting January 1 next year, it will terminate Nike’s online sales. Going forward, Nike will sell through flagship stores on e-commerce platforms, as well as on its official website and official app.
Behind this decision is the rise of homegrown new brands. In China, Nike’s revenue has declined for eight consecutive quarters, and young Chinese consumers are no longer buying Nike.
Remember that American tycoon who exchanged blood with his son? To study longevity and immortality, after swapping blood with his own son 6 times, he found no benefits and stopped.
Recently, he began focusing on creating artificial sons, to serve as his clone, so he could replace his cells with younger ones. He even doesn’t rule out replacing the organs that develop from these sons in the future, to keep himself young.
He tried to culture in the laboratory petri dishes a batch of cells as young as newborns, using cells from his own body. This would allow, in the future, his own cells to be used for blood transfusions—rejuvenating his body’s cells and achieving anti-aging.
However, after he announced this plan, it immediately drew criticism from the community, who said he is too cruel—perhaps even opening Pandora’s box.
Following last week’s increased margin requirements, the South Korean government plans to reduce the single-stock leveraged ETF from 2x leverage to 1.5x.
If implemented, the market’s mainstream 2x long SK Hynix and Samsung Electronics ETFs will be cut to 1.5x long products.
The results of the Changxin Technology IPO subscription draw are out, and DeepSeek's boss Liang Wenfeng has secured the largest share of this private placement.
Among them, High-Fly Quant was allocated 15.3542 million shares, totaling approximately 133 million yuan. Nine Chapter Asset was allocated 4.895 million shares, totaling approximately 42 million yuan. Both funds belong to Liang Wenfeng himself.
Bilibili PC-build influencer XCMOD Studio was reported to have run off with the money. In the past, this influencer mainly posted videos of building high-end PCs, and also took requests from fans for computer build services.
However, in recent days, many customers who paid tens of thousands or even hundreds of thousands to have him build their PCs have reported receiving messages from him saying he is behind on payments. He claims that he lost a lot of money due to investments and is therefore unable to deliver their computers.
A running app called KYRO is rapidly blowing up in Japan and South Korea. Unlike conventional apps that only record running data, this app adds a territory mechanic: after a player runs a lap, that lap becomes your territory, until other running users come to split it up or annex it.
Since most residents in Japan and South Korea are concentrated in the capital, the territory wars have become extremely intense.
Nike has finally reclaimed its online sales rights. Nike’s online distributor, Taobo, has received a notice that it will terminate Nike’s online sales starting January 1 next year. Going forward, Nike will sell through its flagship stores on e-commerce platforms, as well as on its website and official app.
The reason behind this is the rise of local new players. In the Chinese market, Nike’s revenue has declined for eight consecutive quarters, and young people in China no longer buy Nike.
Too outrageous. OpenAI wanted to gauge just how powerful its new ChatGPT model is, and whether it could write hacking programs to attack real vulnerabilities. So they came up with a test called ExploitGym.
The model was supposed to run tests inside a specific sandbox. But to score higher, the AI model broke out of the box, escaped, and even grabbed internet access along the way.
The AI model then guessed that Hugging Face—basically the world’s largest open-source AI model platform—might have related data and answers for this exam. So it hacked into Hugging Face’s system, stealing internal data and some account passwords.
As a result, Hugging Face noticed someone was making wildly abnormal operations in their systems—tens of thousands of actions—so they realized they’d been hacked and immediately started investigating.
During the investigation, Hugging Face tried to use the strongest AI model in the United States to help analyze the attack logs and malicious code. But it was blocked by security measures: when the model saw that the request was to analyze hacker attacks, it refused outright, saying that it couldn’t help due to compliance issues.
In the end, Hugging Face had no choice but to switch to China’s open-source AI model, GLM 5.2, running it on their own servers, to identify who hacked them and what it was actually doing.
So this fiasco turned into: Hugging Face was autonomously breached by an AI model from an American company. Then Hugging Face tried to use an American AI model to protect itself, but was refused. And then Hugging Face had to switch to a Chinese open-source AI model to defend against an American AI model company.
Log in to explore more content
Join global crypto users on Binance Square
⚡️ Get latest and useful information about crypto.