海盗鸭

Gold prices are at 4,000—there isn’t one trending topic post about it.
During the last round of gold price increases, the whole screen was filled with jokes about aunties rushing in to buy the dip.
This time, it’s strangely quiet.
People buying gold aren’t excited this time—they’re hiding.
Hiding from what? Everyone knows.

The weak nonfarm payrolls report in June pushed Bitcoin up to $62,000, with the logic being: bad data = rate-cut expectations = liquidity tailwind. But this Friday’s report could cut the other way—the market mood has already shifted away from "bad news is good news." With inflation still sticky, if employment data disappoints again, traders may worry not about rate cuts but about recession, and in risk-off mode, BTC has a fairly high chance of falling along with risk assets. $BTC Don’t be fooled by the $62,000 level these next few days: the same data can be interpreted in completely opposite ways under different expectation frameworks, and that’s the hardest part of macro trading. Keep a close eye on that half hour during the U.S. session on Friday.

Micron’s shares plunge 39% in a single day, while SK hynix and Samsung roll out $1.3 trillion in capital expenditure plans at the same time—together, these are the real story. The market isn’t afraid of weak demand; quite the opposite. It’s the three Asian giants rapidly ramping up production—pushing HBM supply from tightness toward oversupply—so Micron’s pricing power can’t hold up. This mirrors the same plot as the 2018 memory-cycle peak: the louder the expansion calls get, the more likely you are hitting the profit-margin turning point. The decline of $MUB is already pricing this in early, but the more aggressive the capex from hynix, the more cash flow pressure it will face in the near term, and $SKHYB may not be a safe haven either. The memory sector is in a struggle period before the inventory drawdown—don’t rush to buy yet.

Coinbase’s Armstrong is at it again, saying that AI agents will expand crypto adoption. It sounds like a promise with a shiny pitch, but the logic behind it isn’t entirely hollow. AI agents need to be able to complete payments, handle subscriptions, and call APIs on their own. The traditional credit-card system simply doesn’t fit the high-frequency, machine-to-machine micro-payments—especially when there’s no bank card, you need 24/7 availability, and the amounts could be as small as a few cents. Stablecoins and on-chain settlement are naturally built for exactly this.

That said, wishing for a vision and executing on it are two different things. Armstrong is speaking from a position at $COINB —of course he’d want to tie together the two hottest narratives in crypto and AI and tell a story that benefits his platform. That can be a positive for both trading volume and valuation. The real question is: how many AI agents’ transactions are actually running on-chain, and on which chains? From the looks of it, the settlement layer is the most likely to benefit first. Demand for gas from stablecoins and mainstream chains such as $ETH should show an early reaction. I believe the narrative—at least halfway—but the delivery timeline is likely much slower than what the headlines suggest. Don’t let the six words “AI + Crypto” get you carried away. Treat it as a long-term theme to watch, not a trade that pays off tomorrow.
#AI chips

The detail that Kraken’s security director dug out is pretty painful: the reason the Coldcard bug that was hidden for five years went undiscovered is that the audit only confirmed that a “random number generator exists,” but didn’t verify whether it was actually being used. This is the entire hardware wallet industry’s testing blind spot—people check whether parts are present, but nobody checks whether those parts actually work. Put in the context of self-hosting at $BTC , this is even more terrifying than a certain theft incident, showing that the audit process we trust itself has systematic gaps. Cold wallets aren’t something you buy once and you’re done forever—the security promises from manufacturers also need to be questioned. After this, it’s likely the industry’s audit standards will all need to be rewritten.

Ethereum closed red in July, but it's still 60% away from its all-time high. Meanwhile, network data during the same period has clearly improved—activity and on-chain settlement are both rising. This divergence between “fundamentals improving, while the coin price lags” is a point that anyone discussing $ETH can’t avoid.

Where’s the problem? My understanding is that the value-capture chain has been diluted. L2s have siphoned off a large volume of transactions and users, mainnet gas revenue is being diverted, and the deflationary logic of ETH as an asset isn’t as hard anymore as it used to be. The more prosperous the network becomes, the less fees the mainnet itself receives—this is a structural contradiction embedded in the Rollup roadmap from the start. On top of that, this year’s capital has clearly shown a stronger preference for assets like $SOL —“high-performance single chains with simple narratives.” Something like ETH—“a massive ecosystem but hard to account for”—isn’t as favored by institutions.

So don’t just look at TVL and rising transaction counts and assume ETH should catch up. Between network prosperity and token appreciation, there’s a “value allocation ledger” that sits in between. Until that ledger gets resolved, the divergence can continue. In the short term, I’d rather treat it as range-bound consolidation, waiting for a catalyst that brings fees back to the mainnet.
#Ethereum

A rather unusual piece of data: in 2026, Bitcoin’s volatility is actually lower than South Korea’s KOSPI. In the past, it was always the stock market that joked that crypto was a casino; now it’s Samsung and SK Hynix’s AI narrative that has turned leveraged ETFs into something that makes your heart race. This shows two things—that after BTC became institutionalized, it has indeed been “de-sensationalized” (or “muted”): the influx of big money has diluted the speculative crowd; and the AI narrative has pushed the traditional stock market into the kind of craze that characterized crypto in its early days. The shift from $BTC to “a safe haven” is actually good news for anyone looking to allocate capital—lower volatility means you can take larger positions. Targets like $SKHYB are now the real high-volatility battlegrounds; chasing the price up requires careful weighing.

Galaxy's Alex Thorn warns that Coldcard's alleged fourth wave attack has once again taken away 448 BTC, and this time there's a detail worth noting—some of the stolen funds are still sitting in unconfirmed transactions, giving victims a very narrow window to try to recover them. Hardware wallets have long been treated as a belief in cold storage, but this sequence of four waves has directly punctured the myth of "self-custody is absolutely safe." The root problem lies in flawed randomness generation, meaning the private key could be predictable from the factory. $BTC itself is fine, but this serves as a reminder to everyone using hardware wallets: cold storage doesn’t mean uninspected—your firmware version needs to be checked too. Security is always the most expensive tuition in crypto.

Saylor’s Strategy is now starting to watch Bitcoin’s 200-week moving average, and this move carries a lot of information. The 200-week line has historically been repeatedly respected through every bull-and-bear cycle as long-term support. For someone who never looks at technical indicators and only ever calls out “hold forever” to suddenly publicly focus on this line suggests he is also giving himself an anchor for “should I keep adding or not.” This is the same thing as his pause in DCA—two sides of the same coin: the logic for adding positions has shifted from mindless buying to timing buys. The shareholders of $MSTRB should take note: when the company’s narrative shifts from offense to defense, once $BTC breaks below that line, MSTR’s premium could collapse before the coin price does. That line is now a psychological line of defense shared by two assets.

Saylor has paused the strategy’s weekly fixed-interval investing. That’s a signal. After looking back over the past three or four years, the premium logic of MSTR has always been built on the “infinite accumulation narrative.” When buy-side demand stops, that portion of the story has to be repriced. Now he’s changing his tune to target the BTC 200-week moving average as support, which effectively admits that the current price isn’t cheap and he wants to wait for a pullback before acting. The problem is that MSTR itself has a leveraged characteristic—when the coin price trades sideways, it falls faster. $MSTRB is under short-term pressure. Don’t treat it as a direct substitute for $BTC . As a volatility amplifier, it’s a plus in a bull market and a minus in a choppy market. It’s worth watching the timing of his next entry.

After Apple’s earnings report, it was kicked out of the $5 trillion club—let’s break this down. The market’s immediate concern is memory price hikes—DRAM and NAND have been rising sharply, and it’s true that hardware gross margins are being squeezed. But I think the reaction is a bit overdone. Apple’s bargaining power across the entire supply chain is top-tier. It knows best how to use long-term contracts to lock prices and secure early inventory. In the short term, the cost shock is more about sentiment and valuation cuts than a real collapse in fundamentals.

What’s really worth watching is the other side of the equation: the beneficiaries of rising memory prices. Micron and SK hynix were still cutting production and clearing inventory last year, but this year they’ve turned around. Same event, different outcomes—at Apple it’s cost pressure, while at the suppliers it’s profit leverage. Markets often kill the “visible losers” first and reward the “quiet eaters” later. So instead of obsessing over $AAPLB being kicked out of the club, it’s better to ask who’s actually collecting money in the price-hike chain. Earnings guidance from storage-related names like $MUB may carry more actionable information than Apple’s report. Apple may face pressure in the short term, but that $5 trillion threshold is one it has entered and exited more than once—don’t read too much into it.
#美股财报 #AI chips

Meta This cash flow data is really tough. The ad engine is still running at high speed, and free cash flow hasn’t peaked just because they’re spending heavily on AI and the metaverse—this is the most solid piece in the bull case. It’s not a pure cash-burn story; the core business has been self-generating cash. But I’m focused on another contradiction: on one hand, ad revenue is stable; on the other, capital expenditure guidance keeps being raised. When will this bottomless pit of AI infrastructure finally translate into the revenue side? The market doesn’t have a consensus on that. Zuckerberg’s current approach is to use highly certain advertising cash to bet on an uncertain AI future. The short-term results may look good, but the long-term outcome depends on the payback cycle of this investment. Compared with several other AI-heavy giants, Meta’s advantage is that it has a printing-press machine that’s still making money as a backstop, giving it a much higher margin for error. It’s worth watching the capex inflection point over the next few quarters. $METAB

#US stock earnings report

The XRP ETF has seen net inflows for the fourth consecutive month—money has been steadily pouring in, but the coin price is still about 40% lower than the January peak. This divergence is definitely worth pondering. Typically we assume that institutional inflows mean buy-side support, which should push the price higher, but this time XRP is doing the opposite. My view is that the incremental inflow from the ETF is simply not enough to outweigh the sell pressure coming from earlier holders’ positions and unlocking schedules in the spot market. Funds are entering slowly, while supply is being dumped quickly—so on the books, you get “inflows without the price rising.” This is essentially a reminder: strong-looking ETF data doesn’t automatically translate into price support; between the two sits the entire supply structure. To reverse the situation in the short term $XRP , slow drip-feeding through the ETF isn’t enough—you need to see when the sell pressure finally eases. While tracking fund flows, you also need to look at who’s selling on the other side.

An easily overlooked piece of data: SpaceX's secondary market valuation is already 19% below its IPO pricing. Interestingly, someone dug up Tesla's post-IPO performance — in its early days it also experienced a break below issue price and a long period of consolidation before the later surge. This analogy sounds inspiring, but I have to pour a bucket of cold water on it: history rhymes, it doesn't repeat. When Tesla broke below its issue price, its market cap was only in the tens of billions; SpaceX is now a company worth hundreds of billions, and the same percentage pullback reflects a completely different scale of capital and expectations. Whether this 19% discount is a golden buying opportunity or the start of valuation returning to fundamentals depends on whether Starlink's paying user growth can support that sky-high price. I won't blindly buy the dip $TSLAB or bet on Starlink just because "Tesla was like this back then"; a compelling story and certainty are two different things, and this level is more suitable for watching than for making a heavy bet.

Galaxy research director Alex Thorn pointed out that this Coldcard attack is no small one—possibly the fourth round, with 389 BTC allegedly taken. Even more concerning are the details he mentioned: some of the stolen funds are still sitting in unconfirmed transactions, and victims have a very narrow time window in which they might be able to recover them. This suggests the attack method is targeting the hardware wallet’s firmware or supply-chain component, rather than relying on simple phishing. Many people think a cold wallet is absolutely safe, but as long as there’s contamination in areas like device provenance or firmware updates, going offline won’t be able to block it. The repeated use of the phrase "the Nth wave" indicates the vulnerability may not have been fully patched all along. Users of the same model should really check the source of their firmware. Events like this may have limited direct impact on BTC price, but they slowly erode trust in self-custody—once the security narrative loosens, it can push some people back toward exchanges. $BTC

#BitcoinSecurity

Store this wave—worth bringing up separately. This piece bundles together “the show is still to come” with Nvidia, Micron, and Sandisk, and the logic is actually the same chain: as AI compute expansion grows, it doesn’t just consume GPUs—it’s also疯狂地吞 HBM and enterprise SSDs. Behind each Nvidia card are dozens of GB of high-bandwidth memory. Micron is the most direct beneficiary in the shovel-and-spade market. And Sandisk’s NAND has just gone through a price-increase cycle. What’s interesting is that market sentiment is clearly segmented—Nvidia’s valuation has already been discussed to death, but Micron and the storage sector’s repricing is only just starting. Last year, they were still cutting production due to oversupply; when data-center demand picked up this year, the narrative flipped immediately. I’d rather track the storage side’s catch-up rhythm. The leaders’ positions are already filled, and second-tier names’ upside hasn’t been fully price in yet. Of course, the key premise is whether AI capex can keep going—if it breaks, all three pull back together. $NVDAB $MUB $SNDKB

#AI芯片 # Shortage of Apple chips drags down sales expectations

When Cathie Wood collectively turns cautious on Wall Street, she instead buys more shares of Tesla and SpaceX—a move that, by itself, carries more information than what she says. Cathie Wood’s usual playbook is that when others panic, she steps in. The issue this time is that the backdrop is a bit different—several investment banks have cut their Tesla target prices, citing weak deliveries and Elon Musk’s political distractions. Her logic is still the same: she doesn’t look at the car-selling numbers in the present; she is betting on the long-term option value of autonomous driving plus humanoid robotics. This disagreement is actually classic. Value investors look at the PE ratio and feel it’s priced outrageously high, while growth investors are thinking about what the company will be like ten years from now. Personally, I lean toward the middle—Wood’s judgment has been validated by the AI wave, but her portfolio volatility is also well known. $TSLAB At today’s price, it feels more like a “faith deposit” than a margin of safety. If you dare to follow, you have to be able to weather the drawdowns.

Cramer’s line “Never underestimate this person” was said after SpaceX put its new-generation constellation into orbit, and after Musk hyped Starlink V3 as a “game changer.” The selling point of V3 is that its bandwidth jumps to the terabit level, directly competing with terrestrial fiber-optic networks—this is the real engine behind SpaceX’s valuation, not the rockets themselves, but the network in the sky. Market sentiment was still mocking Musk’s alleged distractions last year, but now that the satellites are actually up there, the tone has changed. In my view, Starlink’s cash flow has begun to support the entire ecosystem, which indirectly validates the narrative around the $TSLAB robot and autonomous driving—same person, same set of engineering capabilities, reused. In the short term, expectations for a Starlink IPO may keep warming up the related stocks, but don’t forget that the hard part is the V3 mass-production ramp. Sending up more satellites doesn’t automatically mean user payments can keep up. The hype is worth watching, but real money is in the installed base.

PayPal's Q2 earnings are out. Total payments reached a historic high of $486.4 billion. It also separately spun up a crypto division—clearly signaling an intention to double down on the stablecoin track. I’m not focused on the pretty headline total, but on the move to create a “dedicated department.” A long-established payments company with 20+ years of experience is upgrading stablecoins from a side experiment into an independent business line. That shows it has decided this is where the next round of payment infrastructure belongs, not just chasing a trend. The real battleground for stablecoins is cross-border transactions and merchant settlement. PayPal already has an existing merchant network in hand—an moat that Web3-only projects in the short term can’t easily take away. Whether this step—$PYPLB —can be turned into actual profit remains to be seen, but the direction is unmistakable. For the roster of stablecoin players, you also need to count the traditional payments giants.

Reddit got hit a little unfairly this round—because Google search traffic started fluctuating, and the market immediately treated it as a major bearish catalyst for Reddit. The logic isn’t hard to follow: a big portion of Reddit’s visibility comes from Google referrals, and it had just signed a major deal with Google for AI data licensing. So when Google search sneezes, the narrative around Reddit’s traffic catches a cold too. But I think there’s a more thought-provoking point here: it suggests that the so-called “AI data goldmine” companies’ lifeline is still actually in the hands of the traffic entry points, and their moat isn’t as deep as everyone is hyping. Conversely, $GOOGLB is the upstream in this food chain—it acts as both the distributor of traffic and the buyer of AI data. When fluctuations transmit to others, it stays firmly in a position to fish from the calm water. In the short term, Reddit’s sentiment will be tougher to deal with, but what really needs watching is the overall gatekeeper variable: Google’s share of search. That’s the key factor that moves a whole range of content stocks.
#美股财报 #Google