One of the most honest pieces of feedback after the AGI Summit:
“\"The stronger the AI, the less we should let it go—when more decisions are handed to AI, what we need to ask is whether problems can be traced back, and whether humans can take over.\"”
This statement precisely hits the core contradiction in the AI Agent track: the higher the level of automation, the more important controllability becomes.
It’s not about not trusting AI—it’s about must keeping a human handoff interface. Otherwise, when something goes wrong, there won’t even be clues to trace what happened.
📌 Trading insight: In the AI + Crypto space, auditability and traceability are key filtering criteria. If a project can demonstrate that AI decisions are transparent and traceable, it has higher long-term value. The risk of pure black-box automation projects is rising.
"It’s not that I think the industry has no future—on the contrary, stablecoins, tokenization, payments, and prediction-market data are still growing. I just feel that many of the things you once did are suddenly no longer needed by the market, and it all feels empty."
The world keeps moving forward, but you’re left behind.
That feeling isn’t unfamiliar. During the mobile internet cycle, anyone who wasn’t at ByteDance, Tencent, or Alibaba probably went through something similar.
At the same time, the security space is becoming even more important. Hackers are getting more powerful with the help of agents, and the responsible security teams are also strengthening their defenses.
The industry is evolving: old roles disappear, new demands emerge. All you can do is keep up with the changes.
📌 Trading outlook: Stablecoins, tokenization, and payments are clear-growth tracks. RWA and payment-related projects are worth long-term attention. The value of the security audit track is rising.
China’s technology industry is becoming a reference point that Silicon Valley cannot afford to ignore.
From open-source large models and biotech to robotics, Chinese companies are reshaping the global tech competition landscape by leveraging lower costs, faster hardware iteration, and tighter industry-chain coordination.
A core insight from Silicon Valley observers: China isn’t copying Silicon Valley—it’s developing an entirely different model for organizing innovation.
Silicon Valley funding = a long-term bet on future growth. Chinese funding = a survival competition under fund terms, buyback provisions, and exit pressure. The cost is compressed long-term R&D, but the outcome is lower costs, faster execution, and stronger commercialization.
Local RMB funds pursue not only financial returns; they also shoulder responsibilities like attracting investment, job creation, and industrial implementation. The companies filtered by a high-pressure environment demonstrate strong price competitiveness and expansion capability once they enter overseas markets.
📌 Deal outlook: China’s tech going global is a long-term trend. Focus on projects related to China’s AI open-source ecosystem and the hardware supply chain. But be mindful of volatility risks brought by short-termism in the venture capital ecosystem.
Why does the Korean market always fall first when global financial crises begin?
Five reasons: 1. Exports account for a high share of GDP 2. The semiconductor cycle is highly sensitive 3. The auto industry depends on global demand 4. Higher foreign ownership—when risk appetite declines, foreign investors sell South Korea first 5. Samsung and SK Hynix have extremely high index weights
Historical patterns: 1997 Asian financial crisis: excessive external debt → spreads globally 2000 dot-com bubble: DRAM prices fall → Korea hits its peak about 3 months early → the NASDAQ drops 78% 2008 subprime crisis: won depreciation → foreign capital exits → KOSPI enters a bear market first → Lehman Brothers collapses in September
The Korean stock market is a leading indicator of the global manufacturing sector and the economic-cycle transition.
📌 Trading takeaway: When KOSPI collapses first, don’t think it’s only a Korean issue. It’s telling you that the global manufacturing cycle is turning. Crypto markets are highly correlated with macro cycles—this is when you should reduce exposure rather than add to positions.
Signature event: Binance’s coin-stock volume has already exceeded that of the cryptocurrency itself.
This week, besides stocks, it also added U.S. Treasury leveraged ETF contracts. Daily trading volume may not be very high—it’s more like an event-driven, high-activity instrument.
But what does this mean? Crypto exchanges are becoming venues for trading traditional financial assets. Coin-stock, and leveraged Treasury ETFs—the boundaries are getting blurred.
For macro direction and hedging, this adds one more option.
📌 Trade view: Binance is bringing in new inflows using coin-stock and traditional financial products, which is beneficial for the BNB ecosystem. However, the “purity” of the crypto market is declining, and volatility from traditional markets will transmit more directly.
Grok 4.5 tops LaurenBench with 56.9%, leading Claude Sonnet 5, GLM 5.2, Claude Opus 5, Kimi K3, and GPT-5.6.
LaurenBench evaluates AI agents’ intelligence in real-world scenarios: conversation, tool use, memory, and safety.
Meanwhile, Grok 4.5 also ranks first on the HighWalk benchmark—assessing an AI agent’s ability to update technical specifications based on code changes. The best balance between quality and operational efficiency.
Elon Musk announced the Grok Build update.
The AI capability race is heating up. Each benchmark’s leaderboard can change hands quickly, showing that no one is truly unbeatable.
📌 Trading outlook: The AI sector’s valuation is high in the short term, but the technical progress is real. AI infrastructure and compute demand are a deterministic trend. In the crypto market, attention is focused on projects in the AI + Crypto track.
At the same time the AI bull market screeches to a halt, money is quietly flowing back to “old-timer stocks.”
This isn’t a coincidence. When AI valuations are pushed into the sky, traditional value stocks become a safe haven instead. History always repeats itself: in the late stage of each cycle, capital rotates from high beta to low beta.
But this time is different—trading volume for Binance-listed coin stocks has already surpassed cryptocurrencies themselves. Stocks are becoming a new narrative in the crypto market.
📌 Trading view: In the short term, funds rotate into risk-avoidance, but the long-term trend for AI hasn’t changed. As the crypto market’s correlation with the U.S. stock market strengthens, rising volume in coin stocks suggests that volatility in traditional markets is transmitting to crypto more directly.
TradeXYZ has decided to fully compensate for all liquidation losses stemming from the Hynix contract pin incident.
Timeline: At 23:01 UTC on July 27, the token marked price for SK Hynix dropped sharply from $1127.9 to $917.25, triggering massive long liquidations. The cause was a real trade in the South Korea pre-market, which was picked up by multiple data providers and fed into an oracle that "operates as designed".
The community estimates liquidation losses at $57 million–$80 million, involving roughly 960 accounts, with the largest address potentially receiving about $2 million in compensation.
The platform acknowledges that the technology worked correctly, but market integrity is the core issue. It also announced an accelerated pricing mechanism reform—reassessing reliance on external venues and increasing the weight of its own order book.
📌 Trading Take: Worth the payout and likes, but this is a "one-time discretionary resolution" and it is not guaranteed for the future. Lesson: In tail-market conditions, an oracle that relies on a single price source is a ticking time bomb.
South Korea’s KOSPI plunges over 12%, KOSDAQ drops over 8%, with circuit breakers triggered—an unprecedented event in history.
In just two days, Samsung + SK Hynix wiped out a combined market value of 53 trillion won. 1.2 million accounts are liquidated, and 360,000 are forcibly closed. Retail deposits shrink by 3 trillion won.
Ironically, SK Hynix’s Q2 operating profit hits a record high, yet it still falls short of market expectations—after a brief rebound of 4%, it falls again by 9% to probe the lows.
Good news isn’t good enough, so it’s bad news. When “beating expectations” becomes the baseline, “missing by not beating expectations” is a disaster.
📌 Trading view: South Korea is a leading indicator of global risk appetite. When the KOSPI first breaks down, it often signals pressure on global risk assets. In the crypto market, be alert to downside correlation in the short term.
Safe processed 130 million transactions in the quarter, a historic high. 54.8 million SAFE tokens have been staked, and on-chain stablecoin supply stands at $6.48 billion.
Meanwhile, Optimism launched OP Enterprise—an institutional custodial chain infrastructure designed to turn Ethereum into an institutional-grade base layer.
The ETH/BTC trend is breaking out. Tom Lee, Chamath, and Lubin are all bullish. The Robinhood chain has sparked heated discussion, with custody, tokenization, and stablecoins all built on Ethereum.
📌 Market call: The ETH ecosystem’s fundamentals continue to improve—long-term bullish. But in the short term, watch for a pullback and confirmation after the ETH/BTC breakout; don’t chase with FOMO.
The industry’s selection criteria are now clear: whoever can bring real-world assets (RWA) into wallets, lending markets, and trading venues so that users can actually use them.
A simpler question about tokens is: when all these activities happen, does the money flow to token holders?
Even if an RWA platform doesn’t issue particularly useful tokens, it can still build a very successful business. The value of tokens doesn’t depend on how great the platform is—it depends on the value distribution mechanism.
This is the core contradiction in the RWA track: platform success ≠ token success. While tokenizing off-chain assets is a trend, whoever can distribute revenue to token holders is the real winner.
Codex can automatically find support and resistance levels & mark high-volume (order-dense) trading zones. AI tools are evolving from “help you write code” to “help you do technical analysis.”
When Codex can automatically identify support/resistance levels and order-dense zones, the barrier to technical analysis is dropping fast. What used to require manually drawing lines and judging actively traded dense areas is now done with a single command.
But there’s a paradox here: if everyone uses AI to draw the same support and resistance levels, does their reliability increase or decrease?
My take: AI increases the efficiency of technical analysis by 10x, but it also drives the alpha of “simple technical analysis” to nearly zero. The real alpha lies in what the AI hasn’t drawn yet—on-chain data, fund flow, and extreme sentiment values.
Tom Lee took an extreme example with Cisco: from 1993 to 2000 it rose by about 100x, and in the middle it went through at least four drawdowns exceeding 40%. The real top didn’t appear until the stock reached 200x P/E and downstream customers started relying on outrageous assumptions to keep purchasing.
Every drop and liquidation is a heap of skeletons—people without patience, rushing to buy the dip, rushing to make quick money.
Money isn’t made by buying and selling. Money is made by sitting there and letting it compound.
Now the storage sector is plunging, South Korea is collapsing, and crypto is grinding lower—panic is at full intensity. But AI demand hasn’t changed, fundamentals haven’t changed; what has changed is leverage and sentiment. On the road to a 100x gain, Cisco pulled back 4 times of 40%+—and each time someone said, “This time is different.” It really is different: every time it falls, it’s leverage that gets wiped out, and the ones that rise are the holders.
My view: patience is the scarcest alpha in this market. Building a BTC “pyramid” position below 60k doesn’t require perfect timing—it requires being able to withstand drawdowns without wavering.
A single $867 trade smashed down Hyperliquid’s positions worth over a hundred million dollars.
What happened: Before the Korea NXT market opened, liquidity was extremely poor. A single order for SK Hynix—only 1 share traded, about $867—instantaneously drove the stock price down by nearly 30%, triggering a trading halt. Hyperliquid’s oracle for the SKHX perpetual contract uses the Korean market’s KRW price to convert to USD. The abnormal price was relayed onto the chain, and SKHX briefly fell 17.9%, liquidating large amounts of highly leveraged long positions.
This isn’t a problem with SK Hynix’s fundamentals. It’s: ultra-thin pre-market liquidity + an abnormally low executed price + direct oracle propagation + on-chain high-leverage concentration = a cross-market cascading liquidation.
Key issue: SKHX is a HIP-3 third-party deployment market. The market deployer is responsible for the oracle source and parameters. Responsibility can’t be attributed to Hyperliquid’s underlying layer alone, nor can it all be blamed on users.
Compensation? If the Korean trade is deemed valid, the oracle will update according to the rules, and the platform will most likely view the liquidation as normal. But if the trade is canceled, the oracle referenced the abnormal price, or abnormal-price protection wasn’t executed, then users have grounds to request compensation.
When a $867 trade can move positions worth over a hundred million dollars, the truly fragile part isn’t the market—it’s the entire price propagation mechanism.
My take: On-chain stocks are fundamentally a price mapping composed of an oracle + a contract + a liquidation system. Trading this kind of asset requires factoring oracle risk into position management—during extremely illiquid periods, price swings may propagate without loss.
Today I sold PUT options worth almost $7 million, most of which are set to be exercised about a month from now.
The logic is very clear: the storage sector’s decline has already pushed sentiment to near its limit. Great traders are all screaming that they can’t imagine a market like this. Storage stocks’ IV is at the highest percentile in history. But the AI revolution is still continuing—demand hasn’t changed. A month ago, the fundamentals were the same—so it’s like the same product is being sold at a 35–50% discount.
Why sell the “hard-to-fill deep out-of-the-money strike” instead of a price closer to the current market? Because I like stable money. Take the MU 660 PUT expiring in one month as an example: selling 15 contracts immediately nets a profit of $47,000. If MU then truly drops below 660 and triggers delivery, the price you’d pay to buy with $1 million would be an extremely good deal.
This isn’t betting on direction—it’s harvesting panic premium. IV being at the highest percentile in history means option pricing is outrageously expensive. What the seller collects is the sentiment premium.
My view: the storage sector’s fundamentals haven’t really broken—what’s down is leverage and sentiment. Selling far-dated deep out-of-the-money PUTs is collecting rent when others are fearful.
Starchild agents can now capture on-chain opportunities or pay for proxy services across 170+ chains via LayerZero, without manual cross-chain transfers.
AI Agent + cross-chain interoperability = the agent is no longer tied to a single chain. This is a real-world use case where crypto and AI converge: letting AI agents manage cross-chain assets themselves.
When agents can independently make cross-chain payments and rebalance positions, the interaction layer of DeFi is being rewritten.
Stargate Finance is integrating NEAR Intents for cross-chain swaps. With its LayerZero roadmap expanding further—cross-chain interoperability is evolving from “works” to “usable.”
When assets across 170+ chains can move seamlessly across chains, liquidity barriers will fade away. The question is: what does this mean for token holders? When users bridge cross-chain, where does the money flow—and whose pockets does it end up in?
The Golden Age of Encrypted Device Accounts Is Here — Multiple open-source, customizable, secure, and easy-to-recover device accounts are now available. This means the barrier to self-hosting is lowering while security is improving.
Hyperliquid cumulative trading volume reached $40.84 billion, with a 24h peak of $5.6 billion. Open interest stands at $3.9 billion, and the daily peak number of independent traders is 60,600. Multiple metrics have hit historical highs.
While BTC and ETH volatility has converged to the floor, perpetual futures exchange data is setting new records. What does that mean?
Liquidity hasn’t disappeared—it’s just migrated. Money has moved from the spot market to the derivatives market, shifting from holding coins and waiting for a breakout to high-leverage speculation. Retail traders aren’t leaving; they’re using higher leverage to bet on even smaller volatility.
ARK Invest says 80% of crypto industry revenue is concentrated in three platforms, and Hyperliquid is one of them. Just look at the futures data to see why—of the Top 15 by volume, 10 are not crypto: SNDK, SKHYNIX, SOXL, MU… It’s pulling traditional market volatility onto-chain.
When crypto’s alpha shifts from L1 tokens to trading infrastructure, all the projects still talking about “our chain is faster and safer” are telling a story that has already ended.
Base’s official roadmap for three goals is out, and the ambition is huge:
1. Trading — put all assets on-chain, run 24/7 2. Payments — stablecoins with global connectivity 3. Smart contracts — become the default chain for AI smart contracts
Put simply, what Base wants to do is to be the “underlying blockchain layer for global finance.” With Coinbase’s user base and compliance moat backing it, plus Base’s speed and cost advantages, this story is far more tangible than what 90% of other L2s are pitching.
While the TVL of other L2s drops back to 2023 levels, Base is one of the few chains still growing. This isn’t a coincidence — a chain with an exchange traffic on-ramp + stablecoin payment use cases + an AI contract narrative has a siphon effect in a bear market.
My take: the L2 culling has already begun, and in the end no more than 3–4 chains will survive. Base is one of the most deterministic survivors. But ETH itself is the biggest beneficiary — because the endgame of all L2s is paying gas fees to ETH.