I came across this randomly today—why isn’t anyone sharing this on my Chinese feed, haha? By the way, who is Jim Cramer? He seems even more powerful than Sun’s contrarian effect—is he the real “oracle” of the stock market? 🤣

I often see terms like “class beauty” and “school beauty” on X, Xiaohongshu, and even in novels. But in reality, consensus around these titles isn’t as concentrated as one might think—it’s actually quite dispersed. I remember before the Gaokao, a few guys in my class were casually chatting during lunch about the prettiest girls in our class. We were surprised to find that each of us had a different girl in mind as the most attractive. In other words, there was general agreement on which girls were considered beautiful—but the title of “class beauty” was split among three or five different girls. And this was just a sample from the boys; if the entire class had voted, the distribution would likely have been even wider. So every time I hear terms like “department beauty” or “school beauty,” I just take it with a grain of salt. A single class can have three or five class beauties; a whole university might have dozens—or even hundreds—of school beauties. Of course, there are exceptions. When someone like Liu Yifei emerges, all other titles naturally fall to her. Back then, in the class next to ours, there was a girl with a celebrity face—very similar to the “milk tea girl” aesthetic—and the boys in her class had extremely concentrated consensus because her beauty was a level above everyone else’s. She was almost certainly the undisputed “grade beauty” across the entire senior year. I suddenly thought of this because a few days ago—I can’t recall which blogger—I saw a Tweet about first-order, second-order, and third-order problems in crypto trading. First-order: retail traders start researching a coin’s fundamentals. Second-order: retail traders start researching how other retail traders view the coin. Third-order: developers or whales study when retail traders will begin analyzing how others perceive the coin—and then manipulate accordingly. First- and second-order dynamics are classic Keynesian beauty contest theory: First-order: Who is the most beautiful? — Retail traders Second-order: Who do people think is the most beautiful? — Smart money / higher-tier players Third-order: When will people start asking “Who do people think is the most beautiful?” — Devs / Whales Third-order devs and whales time their利好 announcements, pump runs, and hype cycles strategically—like posting a photo of a popular girl on the campus bulletin board at just the right moment. First-order looks at Value (fundamentals). Second-order looks at Consensus. Third-order looks at Consensus Acceleration. Ultimately, what we’re trading isn’t just the asset—it’s consensus itself, and the speed at which consensus forms. As I wrote this, it struck me that this is also analogous to options theory: First-order: Spot (Is this girl actually beautiful?) Second-order: Delta (Are more and more people starting to see her as class beauty?) Third-order: Gamma (Is the number of people who believe she’ll become class beauty accelerating?) Given recent market conditions, shouldn’t you sell a coin when Gamma peaks—like Pons? Damn. I feel like I just realized this a little too late.

Before this trip to 2049, I returned to China for several weeks. Since I never stop exercising, every time I go back, I find badminton partners on Xiaohongshu and Xianyu. They’re usually Gen-Z girls—some are professional athletes from top-tier sports schools, others are amateurs who’ve trained for years and reached peak skill. During two-hour singles matches, we chat about everything under the sun, including romantic relationships. And I noticed a relatively common attitude among these girls: Regarding romance—they’re not opposed, but they’re not optimistic. Regarding marriage and children—they’re not interested for now; some are even resistant. This is completely different from how us Gen-Xers thought back then—it feels like a different world. The more I thought about it, the more it reminded me of crypto. 1. Gen-X love was like the crypto market before 2021: Back then, if you committed to someone, you got married, bought a house, and walked the path together. It was just like buying crypto during those years—believe in a narrative, buy in, hold tight, HODL. Sure, there were scams, countless ICOs and altcoins that went to zero, but the beta was huge. BTC and major coins were just a matter of time before they recovered. You basically made money no matter what you bought. Marriage was the same: economic growth, rising wages, soaring housing prices—getting married and having kids felt like going long on the future. Many relationship problems were buried under the tailwind of the times. It’s not that Gen-Xers understood love better—we just happened to be in a bull market. 2. Gen-Z love is like crypto after 2022: LUNA, FTX, high FDV with low circulating supply… after getting rug-pulled so many times, and with Hyperliquid’s rise, people’s first reaction to any narrative is: “Don’t just tell stories—show me the data.” This year, it’s even starting to feel like stock market investing—looking at revenue, users, buybacks… narratives have become just a foot in the door. Gen-Z’s approach to relationships is the same: How much do you earn? What’s your family background? Do your parents have financial burdens? Are you emotionally stable?… Countless relationship influencers have emerged—women follow Ququ, men follow Li Xinye. Men fear becoming exit liquidity; women fear investing years of youth only to get rug-pulled. Everyone’s doing their due diligence, managing position sizes. Many even choose to stay out of the market entirely. Back in our day, phrases like “I’ll succeed one day” or “I’ll love you forever” were big enough to convince many women to invest. Now? Those empty promises don’t work anymore (girls aren’t easy to fool). The current viral “yellow hair and thin muscles” theory feels like Pumpfun-era vibes—when long-term value is hard to build, just grab short-term attention and eyeballs! Beta is gone; now you have to find your own alpha. It’s not that Gen-Z doesn’t believe in love—they just can’t help viewing it through a lens of risk pricing. We Gen-Xers were just lucky—we lived in an era with massive beta. Gen-Z faces a market where they must pick their own alpha. Just like those who entered crypto in 2013 became OGs by lying low; those who entered in 2023 fought hard and ended up as cannon fodder. It’s timing. It’s fate.

Taking advantage of today’s wild ride by Old Ma, let’s fill in the gaps I left a couple days ago about three intriguing AI projects I’ve been following: Jev, Strata, and Atlas. I’ll tie them together using a movie scene—no technical details, just a narrative. If you’re curious, ask your AI or dive into the research yourself. The scene? Terminator. Imagine the T-800 sent back to 1984 with a mission: find and eliminate Sarah Connor. To accomplish this, it needs three core capabilities: understanding spatial environments, making instant decisions, and thinking offline (no SkyNet in 1984). Coincidentally, these three projects map perfectly to Atlas, Jev, and Strata. 1. First, Li Fei-Fei’s Atlas — Understanding Space When Arnold (the T-800) walks into a room, he glances around and instantly knows what’s behind the door, under the table. That’s exactly what Atlas does: given a few images, it predicts what the scene would look like from a new viewpoint. LLMs predict the next word. Video models (like Seedance) generate frames along a timeline. Atlas predicts new perspectives. Many video models now call themselves “world models.” A few months ago, Li Fei-Fei categorized them into three types: renderers, simulators, and planners. A simulator can become a renderer. But a planner must be built on top of a powerful simulator. The T-800 glancing around and knowing what’s in the room? That’s a simulator. Calculating what happens if he jumps or where a bullet will hit? Also a simulator. Deciding whether to chase, hide, or ram a car? That likely requires a planner. Renderers? Barely used. Atlas aims to be the strongest simulator—and eventually evolve into a planner. Its roadmap looks like this: Static 3D space (current) → Dynamic (4D) → Editable → Interactive → Simulator as Planner (endgame) We’re just at the first step of a long journey—but this is essential to building a Terminator. Pure LLMs won’t cut it. 2. Next, Jev — Instant Judgment Remember that scene? The T-800 is in a motel room when the clerk knocks. Instantly, a list of response options pops up in its vision. It picks one—no sentence construction, no thinking. (I think it chose the most rude one…) That’s Jev: no need for an LLM to generate text or follow chain-of-thought reasoning. Just pick from predefined options and output a calibrated confidence score. TypeSafe’s Jev works like this: no text generation. Only selection from fixed options, with calibrated probabilities. Reply with 0.97 confidence → send it immediately. Highest confidence only 0.60 → hold off; defer to a slower, smarter model. In TypeSafe’s demo video, Jev determines whether an invoice is fake. Traditional LLMs take 8 seconds. Jev takes 0.1 seconds. Why so fast? Because it only does multiple-choice—no text generation. Why so accurate? Thanks to RLCD: Reinforcement Learning for Calibrated Decisions. The architecture hasn’t been disclosed yet, but I suspect it’s still built on Transformers—with the generation component entirely removed. Think of it as an ASIC version of an LLM: can’t chat, can’t code—only makes high-frequency decisions. It compresses and simplifies all the traditional LLM decision steps into a single binary or multiple-choice judgment—speed and cost improved by one to two orders of magnitude. It’s like Daniel Kahneman’s “Thinking, Fast and Slow”—Jev is System 1; traditional LLMs are System 2. This opens up an entirely new path for AI beyond generation: decision-making and judgment. If the T-800 relied on today’s LLM speed, it’d be useless in the movie—moving slower than a sloth (suddenly I’m thinking of Zootopia’s Flash…). Jev is the second missing piece of the Terminator puzzle. 3. The final piece — Strata — Running Large Models Locally Yes, the internet existed in 1984—but it was painfully slow and limited to military and academic use. And there was no cloud with SkyNet to connect to. The T-800 had to carry its own model inside its head. Strata enables a consumer-grade GPU with 12GB VRAM and 64GB RAM to run Qwen3.8-Flash-Next, a 125-billion-parameter model. The magic? First, quantize parameters from 16-bit down to 2–3 bits—shrinking size several-fold. Then, keep the most frequently used 20% of experts in VRAM; store all experts in 64GB RAM; finally, store a 30GB lookup table (a training-era cheat sheet) on SSD. Fully offline. Data privacy guaranteed. Sure, performance is reduced due to compression, and reading long documents is slower—but now a high-end gaming PC can run a trillion-parameter model. That’s the leap from nothing to something. We don’t know the exact specs inside the T-800’s brain—but Strata lets you run larger MoE models on constrained hardware (currently only supports Qwen, but the approach is generalizable to all MoE models). Put it all together: we’re getting closer than ever to SkyNet—and the Terminator era. @drfeifei @theworldlabs @typesafeai P.S. If any rich folks have ordered a Roadster—can I ride along in 2028? Please?

WC! WC!!! I just test-drove the Model Y Long Range + FSD last weekend, and it left an unforgettable impression. This week, Elon Musk dropped this massive bombshell—I was completely stunned! Why drive a gasoline car? Why need a fridge, TV, or big sofa? Just drive Tesla!

Last night, upon returning, I saw Ondo’s announcement of Ondo Intelligent Portfolios and suddenly realized: this could be the milestone marking the arrival of RWA 2.0. I’ve repeatedly emphasized that RWA is the biggest theme of this cycle—and what we’ve seen so far can be called RWA 1.0: bringing various assets on-chain—U.S. Treasuries, equities, ETFs, gold… I’d call this the Tokenize Assets era, and Ondo is unquestionably the leader in this phase. But I didn’t expect RWA 2.0 to also be pioneered by Ondo—I’d call this the Tokenize Strategies era. In other words, 1.0 is about putting financial assets on-chain, solving the question: “What assets can we buy?” 2.0 is about putting financial services on-chain, solving the question: “How do we hold and manage these assets?” Take the new Intelligent Portfolios as an example—their structure is actually quite simple: At the base layer, they still consist of already-tokenized assets like Ondo Stocks and ETFs. On top of that, a predefined portfolio strategy determines exactly which assets to buy, their weightings, and when to rebalance. For the initial launch, the first three portfolio strategies were developed by BlackRock for Ondo, while Ondo is responsible for implementing these strategies on-chain—allocating the underlying assets, performing periodic rebalancing, and managing product operations. In the end, you don’t need to hold a dozen different tokens—you simply hold one single Portfolio Token. Ondo has automated the entire process: selecting assets → setting weights → purchasing → continuous monitoring → rebalancing. You might naturally wonder: Isn’t this just like buying a 1.0-era tokenized ETF like QQQ or SPY? At first glance, they seem similar—both are baskets of assets, and neither requires you to manage them manually. But the key difference lies in how they’re built and the cost of creation. In TradFi, launching a new portfolio product requires fund structuring, licensing, custody, and issuance—all with extremely high barriers to entry. As a result, only a handful of “standardized” portfolios exist on the market. On-chain is different. When more and more assets are already tokenized, creating a new portfolio essentially becomes assembling these building blocks via smart contracts: allocation, weightings, rebalancing—all encoded into rules and executed automatically. Every adjustment is transparently recorded on-chain. This is why I believe the true potential of 2.0 lies ahead. The team has already confirmed that future iterations will gradually incorporate Perps, spot crypto, options, and more. At that point, combinations previously nearly impossible in TradFi—like U.S. equities + BTC or commodities + options overlays—will become feasible. (Of course, the current portfolios still use only tokenized stocks and ETFs.) One easily overlooked point: The Portfolio Token itself is a token! A token can be freely transferred between wallets, exchanges, and DeFi protocols. Looking further ahead, an entire portfolio could serve as collateral for lending, margin for Perps, or even be embedded within another portfolio. These are still conceptual directions—but once realized, DeFi’s building blocks will evolve from individual assets to full portfolios. This is the ultimate “financial Lego” vision we imagined during DeFi Summer! Looking back at Ondo’s path over the past few years, it’s remarkably clear: USDY solved dollar yield; Ondo Stocks solved U.S. equities; Perps solved leverage; Network solved infrastructure. With each additional asset tokenized, Intelligent Portfolios gain another possible combination. The more 1.0 building blocks there are, the bigger the houses 2.0 can build. RWA 1.0 gave us ownership of assets; RWA 2.0 gives us the ability to configure them. This was once a service reserved for private banks and high-net-worth clients—now, eligible non-U.S. investors can access it with just their wallets. From Tokenized Assets (1.0) to Tokenized Strategies (2.0)—this is how Wall Street moves on-chain, and this is where RWA truly begins to become fascinating. Thank you to @Ondo for supporting this research.

Thanks to @168MrZ for the invitation to return to 168X—I chatted for over an hour with Z and had a great time. This might be the last “real” bull market; things will likely become increasingly like the U.S. stock market, where tokens without fundamentals or data from “shady VC” projects won’t stand a chance. In other words, during this cycle, I feel many low-quality coins are still being pulled up by the stronger ones, completing history’s final “exit opportunity.”🤨

Last year, a friend went through the dual blow of a liquidation and a breakup, leading to mild depression and somatization. I comforted her by saying: “Apart from life and death, nothing in life is truly major. See yourself as the protagonist of Earth Online, and everyone else as NPCs—including your ex-boyfriend, and even me.” It’s essentially adopting a third-person perspective. Later, she said she tried it—not instantly, but her pain and emotional exhaustion did ease somewhat. Just like Xiao Mao’s comment said: “Go for it!”

StableStock is one of the very few platforms today that allows you to access assets across the U.S., Hong Kong, and Korean markets using USDT—all now available on the mobile app!

I agree with Haotian’s thinking, but I’m just holding both and not making any moves. I bought $SHROOM halfway up the mountain—it feels tough now. But just now it occurred to me: what SHROOM is trying to do, $AI could theoretically do too—and possibly even better. First, AI is already a paired asset for many other memes. If, in the future, these memes deepen their ties with AI while also starting to pair with tokenized small- and mid-cap stocks, AI could effectively become an intermediate node connecting these long-tail assets through multi-hop routing. Of course, this is still just “a path exists”—it doesn’t mean AI itself owns that liquidity. What’s truly more noteworthy is the second point: Beyond NVDA, AI’s Vault/Treasury is already beginning to include other tokenized stocks. If the project team takes one more step in the next update—directly creating pools between AI and stocks other than NVDA, continuously converting capital generated from meme trading into various stock POLs, then reinvesting fees to expand liquidity—it would essentially be walking the exact same path as SHROOM: taking someone else’s route and leaving them with no path at all. Moreover, AI might have an even greater advantage than SHROOM: SHROOM designed a stock liquidity network first, then tried to channel speculative energy from memes into it. AI already has memes—with trading volume and capital formation power—and is now growing a stock liquidity network outward from there. One is building the road first and waiting for traffic; the other already has traffic and is now building the road. If AI truly evolves in this direction, SHROOM’s current goal—a long-tail stock liquidity network—could end up being overtaken by AI.