What does the desktop AI assistant you imagine look like? Is it like this?

#美联储三年来首次加息25个基点 Don't sleep, brothers, get up and chase the rally! Everyone's guess yesterday was right, as expected, the bad news is fully priced in and turns into good news. The Federal Reserve is naturally rebellious and just won't listen to the president; they went ahead and raised rates, but didn't dare to go too far, only a 25 basis point hike, which is within market expectations. The crypto market didn't fall but rose instead, probably blowing up a lot of speculators trying to short last night. The biggest recent bad news, the rate hike, landing is actually a good thing, we can finally catch a bit of a breather. By the way, this $ZEC is really strong, it feels like it's truly Bitcoin now. Such a big coin just surged 20 centimeters straight up. Thinking back to when it dipped below 200 not long ago, why didn't I buy the dip? Sigh.

Liang Wenfeng rumored to have "1.7 trillion liquidation", the truth is more heartbreaking than the rumor These days, a piece of news has been spreading wildly online: Liang Wenfeng's 1.7 trillion liquidation. Don't rush to believe it. This number has no official source, and the management scale of Magic Square Quantitative has never reached the trillion level, let alone 1.7 trillion. It seems more like an exaggerated headline that has been repeatedly forwarded. But the headline is false, the story is true. What really happened is more worth pondering than the rumor. In July 2026, Liang Wenfeng's old business, the quantitative investment empire, suffered a rare major drawdown. In the same month, his new venture DeepSeek's valuation was heading towards 500 billion. The quantitative empire collapsed in one month Let's talk about what really happened. According to data from Private Placement Ranking Network, as of July 31, 2026, among the 9 displayed products under Liang Wenfeng's Magic Square Quantitative, 8 had negative returns for the year, with only one barely holding a positive return of 0.04%, which is almost negligible. In July alone, all 9 products dropped more than 20%, with the core product's maximum drawdown reaching 22.15%. What does this mean? It means if an investor put one million into Magic Square at the beginning of July, by the end of the month, it might only be seventy-something thousand left. In one month, it lost the amount typical of a deep bear market. More heartbreaking is the time span. Over the past five years, Magic Square Quantitative had an average annualized return exceeding 100%, a godlike existence in the industry. The myth built over five years collapsed in one month, and the profits earned in the first half of the year were all given back in one month. The AI empire is expanding wildly In June 2026, DeepSeek completed its first round of external financing since its establishment, raising over 50 billion yuan, with Liang Wenfeng personally contributing 20 billion, making him the largest single investor. Tencent invested 10 billion, CATL system 5 billion, JD.com, NetEase, and IDG each 3 billion, and the national team fund 1 billion. This round of financing set a record in China's AI industry. In early August, Caijing reported that DeepSeek restarted its second round of financing, planning to raise 50 billion yuan with a pre-investment valuation of about 500 billion yuan, aiming to sign the deal in late August. On September 9, news came that DeepSeek had hired CITIC Securities to prepare for a STAR Market IPO, planning to start the listing process within the year. Just a quarter ago, Liang Wenfeng was saying "no financing, no listing," but now financing rounds are coming one after another, and listing is on the agenda. The man who once relied on his own funds to tough it out has finally opened the door to the capital market. Two legs, one is weakening Liang Wenfeng's story has always relied on two legs. One is quantitative. Magic Square Quantitative made huge profits in recent years, and these profits were continuously injected into DeepSeek to buy computing power, support the team, and conduct research. The outside world calls it "quantitative blood transfusion to AI," which was also DeepSeek's biggest confidence to persist without financing. The other is AI. In January 2025, DeepSeek R1 emerged suddenly, wiping out $589 billion of Nvidia's market value in a single day, shaking Silicon Valley. Since then, DeepSeek became China's AI business card and a target for capital pursuit. The drawdown in July caused cracks in the first leg. The safety cushion of quantitative blood transfusion is thinning, and the market is starting to ask: how much money does Liang Wenfeng still have, and is it enough to keep burning? The answer lies in the timeline. While quantitative suffered, DeepSeek's second round of financing started, valuation at 500 billion, IPO preparations followed. This is not a coincidence but an active switch: from relying on his own money to relying on market money. The most ironic comparison Putting the two report cards together, the irony is overwhelming. Among Magic Square's 9 products, 8 lost money this year, all dropped over 20% in July, losing real money and the trust accumulated by countless investors over five years. On DeepSeek's side, valuation is 500 billion, oversubscribed financing, IPO in queue, and Liang Wenfeng's personal net worth rises with the expansion of the AI empire. The same account, two ledgers: the money lost in quantitative is earned back in AI's valuation. Retail investors losing money in Magic Square may be, in another form, part of DeepSeek's valuation. Netizens' jokes are more straightforward: "If you don't make money in A-shares, make money in API; Liang Wenfeng never loses." On August 6, DeepSeek just announced an overall API price increase, charging global developers more "Token tax." While quantitative is bleeding, AI is raising prices, Liang Wenfeng's calculations are sharper than anyone's. Liang Wenfeng's real trouble Of course, switching is not without cost. Financing means dilution, listing means transparency. The Liang Wenfeng who once said "no financing, no roadshows, no listing" will have to accept market pricing, shareholder scrutiny, and quarterly report interrogations once entering the capital market. DeepSeek's past mystery will be thoroughly dismantled by prospectuses and financial reports. A more realistic problem is that AI is a money-eating beast. Hundreds of billions in computing power investment, doubling team size, continuously rising R&D costs, 50 billion financing sounds scary but may only last a few quarters at AI's burn rate. Whether the quantitative leg can recover determines if he still has private funds; whether the AI leg can generate its own cash flow determines if he can remain composed in front of the capital market. Liang Wenfeng, with two legs, has never needed to stand firm on both as much as today. What to watch next Putting aside the rumor of "1.7 trillion liquidation," there are three real observation points. Can quantitative stop the bleeding? Whether Magic Square's July drawdown was due to strategy failure or extreme market conditions, and whether net value can recover in August and September, will decide how much strength remains in the old "quantitative blood transfusion" model. Can financing and IPO land? When will the 50 billion second round financing be officially announced, and whether the STAR Market IPO really starts or remains rumor, once confirmed, DeepSeek's valuation anchor will be officially established. Can the price increase be accepted? After the API price hike, will developers vote with their feet, will competitors seize the opportunity, this is a direct test of commercialization quality. From a small-town youth in Zhanjiang, to a quantitative giant, then to the helm of an AI unicorn, every step of Liang Wenfeng has ridden the wave of the times. Now with one leg weak and one strong, he needs to prove not only how valuable AI is but also whether he can manage two empires simultaneously. Rumors will fade, stories won't. 1.7 trillion is false, but the coexistence of drawdown and valuation in the report card is true.

Doubao, why did you just shut down my device directly?

Thanks for the analysis

It is estimated that this AI desktop will become popular very soon

Yushi Technology, from 444.9 billion to 192.9 billion: a valuation massacre 477 yuan. This was Yushi Technology's closing price on September 11, with its market value officially falling below 200 billion. Looking back 18 trading days, this "first humanoid robot stock" touched 1100 yuan during the STAR Market's call auction phase, corresponding to a market value of 444.9 billion. In 18 days, it evaporated over 240 billion, more than halving. Emotional premium comes fast and goes faster Yushi Technology's issue price was 150.8 yuan, with an issue P/E ratio of 219 times, while the average P/E ratio of the general equipment manufacturing industry it belongs to is only 38 times. On the first day of listing, it opened directly at 1100 yuan, which means an emotional premium of 629% was added on top of the issue price. Then what? Big drops, small drops, slow declines, and then big drops again—a textbook "emotional tide retreat." It hardly gave short-term traders a decent escape window. Growth cliff, profits even more painful In the first half of 2026, revenue was 1.152 billion yuan, a year-on-year increase of 48.54%, which is indeed impressive in traditional manufacturing. But don't forget, its revenue growth in 2025 was 332%. From 332% to 48%, this cliff is too steep. The profit side is even more awkward—net profit attributable to the parent company was 274 million yuan in the first half, indeed turning profitable, but net profit excluding non-recurring items dropped 19.34% year-on-year. The company said this was due to a sharp increase in R&D and sales expenses, which makes sense, but it also reveals a reality: the gap from humanoid robots "being able to run and jump" to "being able to make money" is a commercialization chasm filled with real money. What's worse, even after falling to 477 yuan, the TTM P/E ratio is still above 330 times. Institutional divergence is absurdly large Nomura's target price is 370 yuan, CCB International only 269 yuan, and CITIC is even harsher, with a reasonable stock price range of 130 to 145 yuan. The price range among these three institutions exceeds 200%. The core disagreement boils down to one sentence: Is Yushi a "hardware manufacturing company" or an "AI platform enterprise"? Pricing it as the former means a P/E ratio at most twice the industry average; pricing it as the latter means high valuation during loss periods is tolerable, but the premise is that growth must be continuously validated. When growth falls from 332% to 48%, the market naturally leans toward the former. Small free float becomes an "accelerator" At the beginning of Yushi's listing, the freely tradable shares were very few, with a large number of pre-IPO shares and strategic placement shares locked up, and the earliest large-scale unlocking not until August 2027. Such a tiny free float amplifies both upward and downward moves. On September 11, the turnover rate reached 85.28%—almost all tradable chips turned over in one day. It's not just Yushi's problem; the entire sector is clearing out From early 2026 to early September, the humanoid robot sector index fell 13%, underperforming the Shanghai Composite Index by 12 percentage points. The sector's PE percentile has dropped to 0% in nearly a year, reaching the "cheap" range. But between cheap and worth buying, there is still a key variable: the timeline for performance fulfillment. BOCOM International put it well: the industry's valuation logic is shifting from "technology Beta" to "delivery Alpha." You show technological breakthroughs, but the market no longer pays; you have to present quantifiable orders and replicable delivery capabilities. Yushi's fully autonomous combat technology released on September 7 is indeed impressive, but the stock price still fell the next day. Policy direction is also changing The Ministry of Industry and Information Technology signals a focus on core components and embodied intelligence, promoting robot applications in agriculture, medical care, and elderly care. The National Development and Reform Commission clearly states: prevent blind follow-ups and herd behavior. Translation: industrial support is shifting from "encouraging entry" to "selecting winners." For leading companies like Yushi, this is actually good news—the competitors relying on concept hype will accelerate clearing out. 477 is not the end, but a midpoint Yushi's CEO himself admits that robots doing simple assembly in their own factory still lag in speed and efficiency. This is honest, but it also means: the "humanoid robot general platform" valued at 444.9 billion yuan by the market is still in the early stage of validation. The window for concept premium is closing, and the window for performance fulfillment is opening. This transition will not be gentle.