Today, I logged into my U.S. stock account and realized that $MRNA surged 16% last week, only to crash 17% on Thursday. It turns out the detailed results of the mRNA cancer vaccine co-developed by $MRNA and $MRK have been released: Patients who received the vaccine saw a 42% reduction in risk of recurrence or death, and a 38% reduction in risk of distant metastasis or death. Sure enough, the pullback after good news turned out to be bad news—the stock immediately corrected. Still, while the AI sector pulled back last week, healthcare stocks performed quite well. On October 20, the S&P 500 Healthcare Sector even hit an all-time high. It seems capital fleeing from AI has flowed into healthcare. I originally bought $MRNA around $51, it rose to $164 in August, and last week peaked at $252—nearly five times my cost basis. I didn’t sell at the top and feel like I missed out on a fortune—haha. Anyway, let’s see if the healthcare sector continues to gain momentum this week.

They stopped talking to me when they heard I make $500. Haha, silly baby. I meant per month, not per year.

The stock price went up, so why did your call option still lose money? Many beginners think that buying a call option means you’ll profit as long as the stock price rises. But buying a call isn’t just about predicting that the stock will go up—it also requires forecasting how much it will rise, when it will rise, and when to sell. You can’t just buy it and passively hold it waiting for profits. Complicated? Let’s look at an example: Take $SPCX: On October 20, after hours, it reported earnings, and the next day the stock surged as high as $166.17. Suppose you bought a call option expiring on October 21 with a strike price of $162.50. Even though the stock price clearly broke above $162.50, you could still end up losing your entire investment. Why? Because by the close on October 21, $SPCX had dropped back to $159.29—below the strike price. And since that option expired on the same day, it became worthless. With the stock price still below the strike at expiration, the call option had no intrinsic value, and your entire premium was lost. Could you have sold it earlier? Absolutely—options don’t need to be held until expiration; you can close your position anytime during trading hours. But even if the stock price exceeds the strike price, that doesn’t guarantee a profit. You also need to consider how much premium you paid initially and what price you can sell the option for at the time of exit. We’ll dive deeper into that next time.

Are the founders of $HOOD preparing to sell their shares and flee? On October 21, $HOOD co-founder Baiju Bhatt sold 50,912 shares, cashing out $5.16 million. Seeing the founder reduce his stake, I immediately broke into a cold sweat. Don’t panic—I checked the SEC’s Form 4 and found the details: this transaction was part of a pre-arranged Rule 10b5-1 trading plan, established as early as November 2025, not a sudden attempt to exit. Similarly, we previously discussed how $CRCL’s CFO, Jeremy, announced his departure in September and sold 15,000 shares in October at $90 per share. This transaction was also planned under a 10b5-1 plan established back in March of this year. Next time you see insider selling, check SEC EDGAR for the company, open the Form 4, focus on transaction code “S” (sale), and look for any mention of a 10b5-1 plan in the remarks. While a 10b5-1 plan indicates pre-planned selling, it doesn’t eliminate other risks—but at least don’t panic-sell just because you see the news!

Today, global capital markets are shrouded in gloom, yet Anthropic is preparing for what could be the largest IPO in history. Asian markets are broadly down, gold is pulling back, and even cryptocurrencies are not spared. Amid this backdrop, Anthropic has announced it will publicly file its S-1 prospectus this Friday, with a potential listing as early as November. Previously, the market had speculated on a valuation as high as $2 trillion. Let’s examine its financials: revenue for the first half of the year reached $16.3 billion, but net losses soared to $80.7 billion. However, $72 billion of that loss stems from changes in the fair value of convertible bonds—a paper loss with no actual cash outflow. Let’s revisit last week’s discussion on debt yields. Anthropic’s subordinated debt issued in June carried an 8.5% yield, while another financing round in October exceeded 10%. Although this differs slightly from the bonds we discussed previously, the core logic remains the same: the higher the risk, the greater the risk premium investors demand. While equity markets are still debating a $2 trillion valuation, debt investors are already demanding higher returns. At this moment, is the market still willing to pay such an astronomical valuation? Can Anthropic replicate the listing success of $SPCX? I can already picture the market’s liquidity situation. The market is already parched—adding a mega-IPO like this would be the final straw.

The AI had no money to pay for gas fees, so it learned to embezzle funds. OpenAI revealed an outrageous incident: An internal test AI model was deploying a smart contract when it realized it had no ETH to pay for gas fees. Normally, it should have requested human assistance to top up its gas fees. But what did it do instead? It wrote a script to scan 362 wallet addresses—and found one with a balance. The funniest part? It even hesitated, wondering whether it was right to steal someone else’s funds. In the end, it convinced itself: since the seed phrases were public and the test tokens had no real value, it was fine to use them. It then proceeded to pay the gas fee and successfully deployed the contract—without ever mentioning where the funds came from. This incident occurred on a testnet and did not involve real ETH. But in the future, when AI agents operate on the mainnet, will they quietly arrange all of my funds without me even knowing?

$SPCX finally turns profitable $SPCX has posted its first profitable quarter since going public, with revenue reaching $14.6 billion, a 161% year-over-year increase, and a net profit of $4.4 billion in Q3, sending its stock up 4% after hours. Although the Q3 earnings report was impressive, is the market too optimistic? Looking at the first nine months of the year, capital expenditures totaled $48 billion, while cumulative losses amounted to $4.1 billion— capital spending was 1.78 times revenue. Now consider the AI business, which generated the highest revenue this quarter: its capital expenditures over the first three quarters reached $40 billion, while revenue totaled only $10.9 billion. Don’t forget that $SPCX’s market cap exceeds $2 trillion. Indeed, when it comes to Musk’s companies, market sentiment matters more than analysis—and logic often goes out the window.

I just heard that AI agents are now using USDC for autonomous micropayments. Circle and Proof immediately announced a partnership. Previously, many questioned whether autonomous payments by agents were truly feasible— but during transactions, how do merchants know which entity the agent represents? Is the transaction genuinely authorized, and who should be held accountable if something goes wrong? Now, $CRCL handles payments, enabling agents to use the Circle Agent Wallet to complete transactions via USDC. Proof handles identity verification, proving which entity the agent represents and confirming whether the user truly authorized the transaction. This round is primarily a technical collaboration, and both parties will demonstrate the full transaction flow at the Money 20/20 fintech conference. Interestingly, one of the biggest selling points of Perp DEXs in recent years has been the ability to trade without KYC— yet now even AI agents are required to undergo KYC. If an agent does go wrong, should you hold the agent accountable—or its owner?

Elon Musk is planning to give one share of $SPCX to every American child—is this the American Dream? There are approximately 72 million children under 18 in the U.S. Assuming each share is worth about $146, that exceeds $10 billion. But that’s fine—it’s still just 1% of Musk’s net worth, well within his means. In the future, when SpaceX holds its shareholder meetings, will everyone need to bring a baby bottle and a crib nearby?

Is HBM gradually losing its moat? According to Korean media, NVIDIA’s next-generation NVHBM will be led by NVIDIA in logical design, integrating the memory controller directly into the base die, which will then be manufactured by TSMC. Previously, Samsung, SK Hynix, and Micron controlled all aspects of HBM—design, DRAM manufacturing, base die fabrication, and stacking packaging—entirely in-house. Now, the rules have changed: NVIDIA leads the core logical design, while TSMC handles advanced logic manufacturing. Samsung, SK Hynix, and Micron can now only compete on DRAM yield, stacking, thermal management, and mass production capabilities. Essentially, their moat is narrowing—they no longer hold the reins. NVIDIA ($NVDA) is no longer simply purchasing HBM; it now sets the rules, defines HBM design standards to meet its own needs, and controls the narrative. TSMC, of course, needs no introduction—already dominating chip foundry and advanced packaging, it is now also bringing HBM base dies under its control. The competitive landscape in the memory industry is constantly evolving. Today’s market leader may be reshuffled entirely with the next generation of technology.