After Meta's Muse surged to the top of the free charts, I felt for the first time that AI Agent monetization is no longer just a PPT concept. Muse has started taking over specific tasks like shopping, booking, canceling subscriptions, and negotiating prices, and in the future, it might even make calls and integrate with smart glasses. This change is crucial: chatbots answer questions, but intelligent agents stand at the entry point of "what to buy, where to spend, and how to pay." Whoever controls the purchase intent has the opportunity to earn revenue from subscription fees, transaction commissions, and merchant traffic. What Meta really wants to compete for might not be the next search box, but the step before users make decisions. But I am also a bit cautious. The better the agent, the more it needs to access emails, calendars, addresses, and payment information; the more aggressively it commercializes, the easier it is for recommendations and ads to get mixed together. Whether Muse can make money is not hard to judge; the difficult part is whether it will act on behalf of users or complete conversions for the platform. Between convenience and trust, Meta can only prove itself through its products. #Muse加速扩张，MetaAI投入或迎来变现

Long-term U.S. Treasury yields continue to rise; the biggest pain may not be today's stock prices, but next year's balance sheets. After the 30-year Treasury yield surged to its highest level since 2004, the market is still debating "when it will peak." But the real issue companies face is more practical: the low-interest debt borrowed in recent years is now entering the refinancing window one after another. When old debt matures, interest rates may jump from 3% directly to 6%, and interest expenses will gradually eat into profits, forcing buybacks, mergers, and expansion budgets to be rescheduled. This process won't be as shocking as a flash crash but will last a long time. Especially for companies with average cash flow that rely on external financing, valuations may not collapse first, but operational choices will narrow first. The longer high interest rates persist, the more the market will shift from "telling growth stories" to "checking interest coverage ratios." So, I’m less worried about a single yield spike and more concerned that investors are still pricing companies with the yardstick of the zero-interest-rate era. Rising financing costs will eventually have to be paid by someone. #美债长端利率持续攀升，融资压力升温

In the same week, watching Costco and Micron is much more interesting than focusing on a single earnings report. Costco's quarterly sales grew by 11.2%, with adjusted same-store sales and online business continuing to expand, and earnings per share reaching $6.75. The scariest thing about the membership business is that consumers say it's expensive, but remain honest when renewing. What it sells is not cheap goods, but the certainty of "I won't get ripped off." Next up is Micron for inspection; the market wants to see not just revenue numbers, but whether AI server demand can continue to absorb high-bandwidth storage, and whether supply discipline can maintain prices. These two companies represent two ways of making money: Costco relies on trust to repeatedly collect money, while Micron leverages cycles and technology to amplify profits. The former validates consumer resilience, the latter determines whether the AI market still has depth. What really matters in earnings season is never who beats expectations by a few cents, but whose profits can better withstand sentiment. #财报观察员：好市多业绩超预期，美光接棒

After the rate hike, BTC didn't drop; the most dangerous interpretation is: it has become immune to macro factors. After the Federal Reserve raised rates by 25 basis points, BTC briefly came under pressure but then recovered, appearing indeed resilient. However, this resilience is not mysterious. The market had already proactively deleveraged, with open contracts significantly shrinking and many fragile longs exiting early; after the rate hike, spot ETF funds flowed back in, and capital willing to hold long-term took over the chips. In other words, the market hasn't ignored the bad news, but there are fewer people left to be scared away. What concerns me more is: if real interest rates continue to rise and the dollar strengthens, can BTC still hold the key range? Withstanding one rate hike only proves an improved chip structure, not that it has decoupled from liquidity. True strength means not breaking support when bad news comes and expanding when good news arrives. We can be optimistic now, but it's not yet time to shout "new paradigm." #美联储重启加息，BTC为何仍有韧性？

Trump demands that official U.S. documents rename AI as "superintelligence," pushing the AI regulatory controversy into an even more awkward position with just one sentence. "Superintelligence" originally has a clear meaning in technical discussions, usually referring to systems with capabilities that surpass humans in all aspects. Now, applying this term to all AI is equivalent to mixing chat assistants, autonomous driving, and future extreme risks into one political slogan. The name sounds stronger, making America's leading narrative easier to promote, but regulatory discussions become more prone to losing focus. The federal government worries that too many state rules will slow down business innovation; state governments worry that Washington only cares about competition, with no one addressing employment, privacy, discrimination, and the impact of data centers on local resources. Both sides have real anxieties, but if it comes down to just "development camp" and "security camp" accusing each other, businesses will ultimately face a fragmented patchwork of regulations. Renaming is quick, but responsibility does not disappear because of it. When a model causes an accident, who ultimately bears the consequences—the developer, the deployer, or the user? This question is far more important than what it is called. #特朗普改称超级智能，AI监管分歧升级

OpenAI and Anthropic have successively lowered model prices, which of course makes developers happy, but the first to be squeezed by this round of price wars may not be the model companies, but the AI applications caught in the middle. As the underlying models become cheaper, products that simply wrap a layer of interface and add a few prompts to charge high prices will struggle. Customers will quickly realize that switching to another model or connecting a router for the same functionality can directly cut costs significantly. Without data, workflows, and customer relationship moats, "repackaging" profits will get thinner and thinner. However, cheaper tokens do not necessarily mean a lower total bill. After model price cuts, teams will pack in longer contexts, run more agents, retry more tasks, and in the end, usage may grow faster than the price drops. It's like cloud computing unit prices continuously dropping, but enterprise cloud bills rarely automatically shrink. So now when I look at AI costs, I no longer just compare how much per million tokens costs. I care more about how many calls it takes to complete a real task, how many failures occur, and how much manual rework is needed. Running a cheap model ten times may not save money compared to a costly model getting it right once. #AI模型集体降价，竞争转向成本

Federal Reserve officials keep delivering hawkish remarks one after another, and the market is still asking, "How long will the rate hikes continue?" To be honest, this question might be asked the wrong way. Currently, the Fed is not facing a single inflation factor. Energy prices, fiscal spending, AI data center investments, and stable employment together support demand. As long as the economy does not significantly slow down, officials have room to continue suppressing inflation. What truly determines policy won't be a preset month, but whether high interest rates have cooled demand, wages, and prices simultaneously. What's more troublesome is that if financial markets rally prematurely due to expectations that "rate hikes are ending soon," the wealth effect will stimulate consumption and financing, which in turn weakens the tightening effect. The more eager the market is to celebrate, the less reason the Fed has to ease early. I now prefer to treat high interest rates as an environment rather than a temporary weather event waiting to end. Companies must prove they can profit under expensive capital, and investors must readjust to cash yielding returns and valuations having gravity. This process won't suddenly disappear because of a single dovish statement. #美联储官员密集发声，加息还要持续多久？

Stablecoins are doing something bigger than "crypto payments": putting dollar accounts into phones and then delivering them to places not covered by the U.S. banking system. The Federal Reserve's latest proposal requires regulated payment stablecoins to be fully backed by highly liquid assets such as short-term U.S. Treasuries and establishes an application framework for banks to issue stablecoins. Once the rules are implemented, overseas users holding stablecoins may correspond to greater demand for U.S. dollar assets and U.S. Treasuries. For users in regions with high inflation, capital controls, or weak banking services, they may not care about on-chain governance; they just want a dollar tool that doesn't rapidly depreciate and can transfer funds even on weekends. The lower the usage threshold, the easier it is for the dollar to bypass traditional banks and spread outward. This is somewhat ironic. Many people think cryptocurrencies will weaken the dollar, but one of the fastest-growing applications is actually helping the dollar gain internet-level distribution capability. Stablecoins may disrupt banks and card networks, but they do not necessarily challenge dollar hegemony; instead, they may become the fiercest new channel for the dollar to go global. #美元稳定币或加速出海

The yield on Japan's 10-year government bonds has risen to about 3.055%, reaching a new high since 1996. For those accustomed to Japan's zero interest rates, this figure even looks somewhat unfamiliar. The danger is not only that Japanese bondholders are losing money. Over the past few decades, many global trades have been based on a simple premise: financing with cheap yen to buy U.S. Treasuries, U.S. stocks, or other high-yield assets. When Japanese interest rates rise and the yen may rebound, the profits from this trade thin out, and some funds can only reduce positions and return home. This is also why bond volatility in Tokyo can transmit to New York. Japanese institutions are important global buyers of overseas assets; when domestic bonds finally offer decent yields, their motivation to continue bearing currency risk and traveling abroad to buy bonds decreases. The U.S. and Europe, wanting to maintain low financing costs, will also lose a stable buyer. Normalization of Japanese interest rates sounds like a domestic policy but is actually tugging at global capital flows. When the faucet is turned down just a bit, highly leveraged assets far away may be the first to feel thirsty. #日本10年期国债收益率创30年新高