The bond market situation is so messy that even Bassett admits it can't be stabilized. Liquidity is so poor that the Treasury Secretary himself said no equilibrium price can be set; increasing repo operations is just a stopgap, not a cure. The pricing power of long-term interest rates is slipping away from the Fed, and whoever holds duration can't sleep well. This will inevitably lead the market to reprice sooner or later. Saving this for now, will review tomorrow.

1.93 trillion evaporated in one day, Nasdaq dropped over 1000 points, officials said they have never seen this level during trading. If you have stocks, don't rush to buy now, first see if the market can hold up at tomorrow's opening. What color is your brokerage app showing? Mine is dark green.

Italy's PPI jumped from 7.8% to 10.9%. Normally, no one pays attention to this data, but once the market acknowledges it, it's a different story. Don't rush to conclusions; first watch if European yields and the euro move in the same direction—only when both move together does it count. Check the results again over the weekend and compare then. Keep an eye on it.

In 1985, nuclear power accounted for more than half of Taiwan's energy, but now it has dropped to zero, and the first step to restart it has just been taken. Energy issues have cycles longer than elections; those who shut it down know this well. The anti-nuclear stance has been played out, and electricity prices and TSMC are the first to bear the burden. How far do you think this step can go? power

Everyone in the group is shouting to go long, but I choose to go against the trend. Trump wants to convert the White House briefing room back into Roosevelt's indoor swimming pool — that's a bold move. Asian session liquidity is poor, and unreasonable prices often appear; don't follow the crowd at times like this. For additional comments, see the comment section.

The S&P 500 is just 0.7% away from its peak, but 430 stocks are already nearly 22% below their highs. This scene is too familiar; it played out the same way in '73 and '99, and the index was directly cut in half afterward. I'm just a retail investor, so I don't dare to guess the top, but every time before a Fed decision, the market moves ahead of expectations. This time, with breadth this bad and the index still holding up, it's obvious who's carrying the weight. No rush to draw conclusions yet; let's watch as it unfolds.

$SPX 2026 The biggest trap is right here: The Fed is still stubbornly tightening, tech giants have already started to crack, and retail investors keep rushing in with every dip. The 2018 script is now replaying. Don't ask why; the answer is that until liquidity returns, every rebound is just for you to sell. For additional comments, see the comment section.

I've always thought that tariffs themselves do not equal a trade war. Tariffs, quotas, anti-dumping, safeguard measures—these are all common tools in international trade. What truly turns the situation into a "war" is often when, right after the negotiation table breaks up, one side turns to face the cameras and packages itself as the bullied victim. Everyone protects their own industries— the US, the EU, Japan, China, India—each has its own subsidies, procurement preferences, and tariff barriers. The difference lies in that some quietly do it, hiding industrial policies within technical standards, government procurement, and R&D subsidies; while others insist on playing the victim, outsourcing domestic distribution problems to an external opponent. This kind of drama has happened many times in history. In 2002, the Bush administration imposed safeguard tariffs of up to 30% on imported steel, claiming to save the US steel industry. The EU, UK, and others immediately retaliated, threatening tariffs on politically sensitive goods like Florida orange juice and South Carolina textiles, and filed complaints with the WTO. The WTO ultimately ruled the US was in violation. Domestic downstream manufacturers also complained about rising steel prices, with increased costs for automobiles, home appliances, and machinery. By the end of 2003, the US itself canceled the tariffs. Summing up, the US steel industry did not revive because of this; downstream industries ended up paying more costs, and the supply chain detoured where it needed to—purchasing from elsewhere continued, and layouts in Mexico, Canada, and Southeast Asia did not stop. Later in 2018, the Trump administration imposed 25% tariffs on steel and 10% on aluminum under "Section 232," with the EU, Canada, Mexico, and others retaliating. The story was similar: steel prices spiked short-term, US steel companies enjoyed profits briefly, but downstream costs rose, trade partners detoured to third countries, and global supply chains were reassembled. Studies estimate that such tariff protections brought limited employment benefits to the steel industry but imposed higher costs on downstream manufacturing, even dragging down GDP and employment in the long term, potentially reducing hundreds of thousands of jobs. In other words, wielding the tariff stick loudly does not necessarily mean winning; it is more a political posture. This current round seems more driven by votes than economic logic. Steel, aluminum, automobiles, and chips are often concentrated in a few swing states or key districts. Politicians need to show rust belt workers "I am protecting you," even if the overall industrial chain is not cost-effective. The US steel industry directly employs only a few hundred thousand people, but the downstream steel-using industries number in the millions. When tariffs rise, steel mills may benefit short-term, but automobiles, machinery, and construction face more expensive raw materials. Tariffs can quickly make headlines, more "visible" than subsidies, retraining, or infrastructure investments. But economic logic asks: where is the comparative advantage, who bears downstream costs, will allies retaliate, will supply chains detour? Vote logic asks: who looks more like a victim in front of the camera, who can reduce complex issues to "they are taking advantage of us." That's why I say tariffs themselves do not count as a trade war. The real fight starts when, after negotiations, one side turns to the camera and cries victimhood. Everyone protects industries; some do it quietly, some insist on playing the victim. When the US and UK tangled over steel tariffs back then, neither gained an advantage; supply chains detoured as needed. This current round is more vote-driven, not economically driven. If you have different views, please share your reasons.

CPI looks better after excluding rent, but the pain felt by ordinary people won't disappear. Inflation is such that good core data doesn't mean living costs have decreased; food, energy, and rent are the real expenses paid out. Data is for the market to see, bills are paid by oneself.

RISC-V wanting to be compatible with CUDA sounds off right from the start. It's not that the open-source instruction set lacks ambition, but what it’s trying to tackle is the wall that Jensen Huang has spent nearly twenty years building with countless lines of code and developer habits. Since CUDA launched in 2006, nearly two decades ago, NVIDIA officially claims millions of CUDA developers, with layers upon layers of ecosystem components like cuDNN, TensorRT, NCCL, OptiX, and Omniverse built around it. The real moat has never been a single chip, but this entire ecosystem: if you change the hardware, the code, libraries, tuning experience, and deployment pipelines might all have to be redone. Ask anyone working in 3D rendering—Arnold, Octane, Redshift—these three renderers are basically all tied to NVIDIA cards. Arnold’s GPU rendering depends on NVIDIA’s CUDA/OptiX; OctaneRender has long been CUDA-centric; Redshift, although it later supported Apple Metal, still can’t avoid NVIDIA in mainstream workflows. It’s not that they don’t want to support others, but they have no choice—supporting AMD, Intel, or RISC-V means rewriting backends, redoing optimizations, and readapting a bunch of plugins, which is commercially unfeasible. So RISC-V’s call for CUDA compatibility feels more like a gesture. To truly make it work, it’s not just about translating a few instructions; you have to fully adapt PTX, SASS, drivers, compilers, math libraries, and frameworks. The open-source community has tried—for example, ZLUDA attempted to run CUDA programs on non-NVIDIA cards, but compatibility, performance, and maintenance costs were all problematic. The community’s goodwill alone can’t fill this gap. Liquidity in the contract market is lowest at dawn, and it’s also when spikes are easiest to occur—similarly, the thickest ecological barriers are never broken overnight. CUDA’s lock-in effect is the result of twenty years of rolling accumulation. For RISC-V to bypass it, it first has to answer why developers would migrate, who bears the performance loss after migration, and what happens to existing code. Maybe I’m wrong and waiting to be proven otherwise.