资管舵

On the morning of September 26, the University of International Business and Economics released a notice of condolence to the public: the renowned musician Liu Huan passed away in Shanghai at 9:52 a.m. on September 25, 2026, at the age of 63.
Liu Huan was born in Tianjin in August 1963. In July 1991, he joined the University of International Business and Economics. He served as a teacher and then as an associate professor in the Arts Teaching and Research Office, and retired in September 2023. He taught for more than three decades. The (History of Western Music) course he lectured on is one of the school’s classic aesthetic-education courses, and his classes were always fully attended. He also composed the school song (UIBE Song), whose melodies are repeatedly sung at the school’s major ceremonies.
The school said that he combined music education, music composition, and singing in one person; he was well-versed in both Chinese and Western cultures, with high professional ethics. He devoted his entire life to the music事业. His passing is a major loss to the school and to the Chinese music community. After leaving campus, he was even better known as a singer to every household; works such as (The Song of the Hero) and (Me and You) are still widely sung today. In 2011, Liu Huan was appointed as a director of the school’s third board of directors.

Bill Gates has been speaking very bluntly these past couple of days: that artificial intelligence is powerful enough to trigger an event that could kill 1 billion people. The remarks come from an interview clip from a program he is set to air this weekend. His exact words were that historically there has never been a weapon that is as terrifying as "malicious people" paired with the latest AI tools.
These remarks aren’t coming from just him. Earlier this month, Jacob Coxon, a researcher at Anthropic, simply resigned, leaving behind the line: "current AI development is a high-stakes gamble with the lives of all humanity." The alignment science lead at the same organization, Evan Hubinger, was even more direct—within the next 10 years, the probability that AI leads to human extinction will exceed 10%. Warnings from Silicon Valley about AI’s potentially catastrophic consequences are being ratcheted up as the technology advances rapidly.

Meta’s Muse has sparked a buzz in both the AI community and the market this week, seen as the next narrative starting point for the expansion of AI demand—"AI demand driven by products and scenarios." At Meta Connect, nearly all newly announced AI hardware revolves around this AI agent. Zuckerberg has declared, "The future belongs to everyone," and at the core is Muse. The expected endgame is a super-intelligent platform reaching a scale of billions of users.
Muse bundles browser operations, account connections, payments, and backend execution together, greatly lowering the barrier for ordinary users to use AI agents. Daily active usage data crushes ChatGPT’s performance in the same period, and the rapid expansion of the user base further confirms it. The biggest difference from earlier AI agents is this: behind every user runs an independent cloud-based virtual machine. Even if the app is closed, tasks continue executing in the background.

According to data from the trade intelligence firm Kpler, despite ongoing escalation of regional conflicts, Saudi Arabia’s crude oil export volume this month still surged sharply. This out-of-trend performance has drawn attention from the market.
Saudi Arabia’s daily crude oil exports in September reached 6 million barrels, the highest level in about seven months, with export volumes having returned to the monthly average level of 2025. Notably, this month Saudi Arabia was forced to shut down a key eastbound oil pipeline after a drone attack, yet the country still managed to increase its crude oil exports—September exports jumped by nearly 80% from August’s low of 3.4 million barrels per day. This indicates that export volumes previously held back due to supply disruptions are being rapidly backfilled.

The wind the Fed is blowing out this week is hawkish enough.
Earlier this week, it first raised the target interest rate range by 25 basis points—up to between 3.75% and 4%—the first rate hike in three years. Since this hike took effect, many Fed officials have publicly called for further tightening; Friday was when they all burst forth at once. Next, just on Friday, three regional Fed presidents came out one after another with messages that were highly consistent: inflation hasn’t been brought down yet, and policy still needs to stay tightening.
Hammack in Cleveland was the most blunt. She worries that high prices will drag on long enough for ordinary people to slowly come to treat inflation as the norm—which is exactly the situation the Fed cannot accept. She said this during an event hosted by the Fed in her district. Her exact words were: "The biggest risk now is that an inflation mindset is starting to take shape." She stressed that the Fed must play a role to ensure policy remains in a restrictive stance and to bring inflation back to the 2% target. Hammack also noted that the economy itself is still growing well and the labor market remains stable, but she is not comfortable with the price pressures coming from the demand side—especially since capital expenditures over the coming period will continue to pose pressure on inflation. She also issued a warning: if progress in bringing inflation down stalls, inflation expectations could shift. As for how AI will affect prices in the long run, she acknowledged that there are still many unresolved long-term questions at present.

Apple has landed in a major patent lawsuit. A federal jury in San Diego, USA, ruled that Apple must pay more than $5.7 billion for infringing two tactile feedback patents owned by Taction Technology. The technology at issue is a low-frequency vibration tactile transducer that can be perceived through touch, and it is used in multiple iPhone and Apple Watch models. The jury found Apple liable for infringement, but it did not find willful infringement, which kept punitive damages out of the decision. Taction formally sued Apple as early as 2021. For Apple, which has deep cash reserves, the money isn’t likely to hurt in the short term—but patent wars are ongoing invisible costs: with every technical iteration, companies have to first work out the patent bills of others. A tech company’s moat is never written only in its own code.

All three major U.S. stock indexes closed higher on Friday: the S&P 500 rose 0.51% to 7,743.41, the Nasdaq gained 0.48% to 27,068.72, and the Dow increased 0.93% to 51,828.62. All three indexes finished the week in the green. The Nasdaq rose 2.06% this week and hit a fresh closing high, the S&P 500 gained 1.21%, and the Dow rose 0.28%.
U.S. Treasury yields surged and then pulled back, with larger declines in the short end: the two-year yield fell about 7 basis points to 4.86%, the ten-year yield dropped about 4 basis points to 5.17%, and the 30-year yield continued to make fresh highs since 2004, still trading around 5.5% and moving in a tight range.
Philadelphia Semiconductor Index rose 1.41% Friday, up 6.27% for the week so far, marking a four-straight-week gain and the longest weekly winning streak since May 5. Apple gained 1.53% to set a record high closing price; Microsoft rose 3.66%, adding roughly $135 billion in market value in a single day, boosted by the release of the new Copilot. Meta fell 3.33% after it was reported that its star AI agent, Muse, faced a computing bottleneck and that services were downgraded, though the stock still ended the week up 12.99% overall. Nvidia edged up 0.22% and Broadcom rose 0.7%. Tesla slipped 1.54%, while SK hynix rose nearly 2.78%.

In the past few months, Tesla has increased the production of its Optimus humanoid robots by about tenfold. It sounds like a mass-production inflection point has arrived, but people familiar with the project say the company is still far from large-scale, reliable production. Last month, Tesla built only a few hundred robots per week. Between that and Musk’s blueprint of producing one million units per year lies an entire supply chain.
The bottlenecks are very specific. First, the robotic hand is too complex—dexterous hands are among the hardest pieces of hardware to crack for humanoid robots, with many joints and high precision requirements. Second, the automated production line hasn’t been running smoothly; many steps still rely on manual debugging. Third, suppliers can’t keep up—yields and delivery schedules for key components limit the overall production cycle time.