小楼

Even in a bear market, there can still be small opportunities—the key is whether your wallet can support quickly.

On-chain and in the secondary market have started to pick up a bit:
• Robinhood chain goes live; OKX wallet supports it right away
• A new UNI platform is launched; OKX wallet follows suit
• Having small opportunities in a bear market is a good thing

All exchanges are laying off staff, and developers are still working—that’s a blessing. An exchange wallet’s rapid support for a new chain means you can capture liquidity and airdrop opportunities at the earliest time. In a bear market, this kind of agility is alpha itself—while others are still queuing to top up, you’re already interacting on-chain.

$BTC #OKX #DeFi

Trading view: Don’t chase highs for small bear-market opportunities. But watch for opportunities to get wallet access to a new chain as soon as it launches. Don’t move BTC spot holdings for the long term.

Store “Three Families, Same Class, Different Fates”: Micron is the top performer; SK hynix is at the bottom.

Hyperliquid’s data from the past 30 days:
• Micron (MU) fell 5.7%—the least
• Sandisk (SNDK) fell 23.8%
• SK hynix (SKHX) fell 31.8%—the most
• MU outperformed SNDK by 18.1 percentage points, and outperformed SKHX by 26.1 percentage points

In the same sector and the same drawdown, the strength gap is 26 percentage points. What does this mean? In a declining market, a pair trade that goes long the stronger and short the weaker is safer than a one-way trade. On-chain perpetual contracts make this strategy easy to access—you don’t need to open two brokerage accounts; one wallet is enough.

$BTC #Hyperliquid #AI

Trading view: Relative strength in the storage chain is one of the most effective ways to trade right now. Avoid one-way positions; focus on pair trade opportunities.

Vitalik personally praised MiniMax H3: the first open-source video model to surpass Tencent HunyuanVideo 1.5.

MiniMax H3 specifications:
• Open-sourced in August 2026
• Supports unified input of text/image/video/audio
• Can generate 15-second 2K videos with native stereo sound
• Leading performance in video editing benchmark rankings

An ETH co-founder personally gave the AI video model a thumbs-up—this is not a simple resharing. Progress in open-source AI video models means content creation costs will keep falling. The explosive growth in the supply of AI-generated content is the underlying driving force behind the web3 creator economy—NFTs, content provenance/verification, and creator tokens all require low-cost content production.

$ETH #AI #Open-source

Trade outlook: The ETH co-founder is paying attention to the AI video track, and the AI × Crypto narrative continues to gain momentum. Hold ETH long-term and keep an eye on web3 projects focused on AI content creation.

Putting Liang Wenfeng’s investment map together is quite interesting:

Storage — Hoshino/9th Chapter’s allocation for Changxin 1.75 billion; first-day unrealized profit of 827 million
AGI — Personal investment in DeepSeek of about 20 billion, accounting for nearly 40% of the funding; the largest single investor
Robotics — DeepSeek received an allocation of 140 million for Unitree robots, with a 36-month lock-up period

From Changxin’s DRAM to DeepSeek’s AGI, and then to Unitree’s humanoid robots. If AGI ultimately enters the real world, it won’t just exist in a chat box. Compute → reasoning → embodiment: this is a complete AI deployment path. Liang Wenfeng isn’t investing in three companies—he’s investing in the full industrial chain of AGI.

$BTC #AI #AGI

Trade thesis: Investment acceleration across the entire AI infrastructure industrial chain. Keep holding BTC spot; continue to watch the AI narrative as a macro backdrop.

AI Agents shine on the OP mainnet.

MetaMask Agent wallet launches on the OP mainnet:
• Provides a new on-chain transaction route for Agents
• Median transaction fees are only $0.00001
• Even with dozens of transactions, execution costs remain extremely low

Agent + a low-fee blockchain = paradise for automated trading. AI Agents can execute strategies on-chain automatically without any manual intervention, and the $0.00001 gas fee makes high-frequency strategies possible. This is the most grounded piece of the AI × Crypto narrative—not a PPT, it’s actually running.

$OP #Optimism #AI

Trading takeaway: The launch of a low-fee chain for Agent wallets is the most practical scenario for AI × Crypto. The OP ecosystem Agent track is worth keeping an eye on.

Opinion moved the BTC/ETH/BNB hourly up/down contracts to a Chainlink CRE workflow.

Details:
• Built on Chainlink’s reliable oracle data
• Automated settlement with zero incidents and zero disputes
• Seamless migration and continuous operation

Oracles are the water, electricity, and gas of DeFi. Without reliable pricing feeds, derivative contracts are just castles in the air. Getting the CRE workflow running means on-chain derivatives’ settlement infrastructure has taken another step forward. In DeFi, the phrase "zero incidents and zero disputes" is worth more than anything else.

$LINK #Oracle #DeFi

Trade assessment: Oracle infrastructure maturity is the prerequisite for a boom in DeFi derivatives. The LINK infrastructure keeps iterating—no rush for the short term; observe ecosystem adoption rates.

Jesse Pollak once said a bold claim: "Every financial institution worldwide will move its operations onto the crypto rails."

The data is actually solid:
• Base TVL has grown by 35% since 2025, while all other ecosystems are declining over the same period
• Base is the world’s second-largest lending market
• Spot trading volume of BTC and ETH on Base exceeds that of any other chain
• On x402, 90% of proxy financial transactions occur on Base

L2 is no longer Ethereum’s sidekick. When traditional finance starts migrating to on-chain rails, low-fee, high-throughput L2s are the first stop. Base’s TVL growth against the trend isn’t a coincidence—it’s not taking share from other L2s, but capturing the entry point for traditional finance.

$ETH $BTC #Base #DeFi

Trade outlook: Base ecosystem TVL is growing against the trend by 35%, and the flow of capital toward L2 is clear. But ETH itself is still finding its footing—don’t chase; wait for confirmation after a breakout.

Observations on the U.S. stock earnings season: with the exception of Amazon and Microsoft, which are “into the cloud,” almost no one else reports earnings and doesn’t drop. No matter whether the results are good or bad, it falls.

This is called “selling facts.” The market has already priced in the good news; even if the earnings are strong, they’re only confirming expectations. On the other hand, cloud providers can go against the trend because AI computing demand follows an incremental logic.

High U.S. stock valuations + uncertainty about interest rates = the market doesn’t buy good news.

$BTC
#美股 #财报季 #macroeconomics

Trading view: The U.S. stocks trading range at elevated levels puts short-term pressure on BTC, but it doesn’t change the long-term DCA plan.

Exchange vs On-Chain High Reward-to-Risk Pair Comparison:

Exchange: enter with 1 unit, target 1.25, stop loss 0.95, reward-to-risk ratio 5.0
On-chain: enter with 2 million FDV, target 20 million FDV, stop loss to zero, reward-to-risk ratio 10.0

The on-chain reward-to-risk ratio is higher, but the trade-off is the risk of going to zero. The exchange’s reward-to-risk ratio is lower, but it offers stop-loss protection.

There is no right or wrong—only the ability to match risk management. The other side of a 10x on-chain reward-to-risk ratio is 100% principal risk.

$BTC
#交易策略 #链上 #Risk management

Trade decision: Don’t touch on-chain shitcoins. No matter how high the reward-to-risk ratio is, going to zero is still going to zero.

Thinking Machines is also an early supercomputing + AI company.

Looking at the history of AI development, you’ll find a pattern: the entanglement between supercomputers and AI didn’t begin today. From Thinking Machines to Nvidia, compute power has always been the physical foundation of AI.

History won’t repeat, but it will rhyme. The starting point of every new wave of AI is a breakthrough in compute infrastructure.

$BTC
#AI #历史 #Compute power

Trading judgment: no trading, but the compute narrative is the long-term main line of the AI cycle.

Full-Stack Programming AI Has Finally Reached a Usable State:

- gpt-5.6-sol ultra: Complex full-stack programming is usable—definitely turn on ultra
- Claude Fabel 5: Slightly stronger overall capabilities; high is enough
- The Fabel 5 safety guardrails are no longer as strict—when you hit limits, switch to opus 5 ultra code

AI programming capability is crossing the threshold from “it works” to truly “it can be used.” The leap from “help you write a function” to “help you build a full stack” is a qualitative change. Developer productivity will increase exponentially—yet this also means the value of junior developers will quickly become obsolete.

$BTC
#AI #编程 #Large Model

Trading judgment: Not trading directly, but the AI programming capability breakthrough is a long-term positive for the entire tech sector.

Ethereum core developers propose introducing faster slots in the Hegotá upgrade.

Core logic: faster slots = greater block space value = benefits users, the network, and especially ETH.

This is a one-time cost in exchange for long-term gains. It is the prerequisite for the roadmap toward fast finality, post-quantum security, and zero-knowledge proof technology. Hegotá may be the last reasonable opportunity to pay this cost for a very long time.

Ethereum’s underlying layer continues to be optimized, and the moat for long-term holders is widening.

$ETH
#以太坊 #技术升级 #Hegotá

Trade thesis: bullish for ETH long-term value—maintain DCA and don’t touch it.

Polaris documentation release brings pETH and its steadily rising floor price, ushering in a new on-chain foundational component ecosystem:

- Next-generation yield-bearing assets
- ETH loans that cannot be forcibly liquidated

ETH financial primitives are evolving. From staking to yield-bearing to non-liquidatable loans, the capital efficiency of on-chain ETH is continuously improving. Polaris may be the next ETH infrastructure project worth keeping an eye on.

$ETH
#DeFi #ETH #Polaris

Trade outlook: The ETH infrastructure narrative is accelerating. Keep an eye on the pETH floor-price mechanism, but don’t chase early moves.

DeepSeek announcement: plans to raise API pricing overall in the near term, with an expected increase that will be relatively significant.

"The golden age is over"—but a 70% rise is still cheaper than what others charge.

DeepSeek’s price increase indicates two things:
1. AI inference costs are bottoming out, and the era of free lunches is ending
2. The pricing power of domestic large-scale models is strengthening

For users, it means higher costs; for the industry, it means moving toward maturity.

$BTC
#AI #DeepSeek #Large model

Trading outlook: I won’t trade directly, but the AI cost curve bottoming out is a positive signal for the compute-power sector.

CryptoRank Statistics: 1,539 tokens have historically entered the top 100 by market cap, and 1,107 have already been in "operational death," accounting for 71.9%.

Death rate curve:
- 1 year: 29.2%
- 3 years: 43.4%
- 5 years: 61.7%
- 10 years: 84.7%
- 12 years: 91.5%

The median lifespan is only 2 years and 4 months. Even harsher: entering the top 100 has not changed the fate of most projects.

The "once-in-the-top-100" coins you hold are most likely already corpses. Treasure the tokens that are still on the list.

$BTC $ETH
#市场分析 #加密市场 #Investment

Trade takeaway: This data once again confirms the correctness of the strategy "hold only BTC + ETH." Altcoins are for spending, not investing.

The next bottleneck for AI may be water.

In the United States, data centers directly consumed about 17 billion gallons of water in 2023, and that could rise to 2–4 times by 2028. Large data centers can use up to 5 million gallons of water per day. In some server rooms, more than half of the water comes from municipal drinking water—AI’s “cloud” is essentially drawing from the city’s tap.

Arizona has rejected an Amazon data center, and Kansas City requires approvals from the city council for new projects.

The hidden costs of the computing-power race are starting to surface.

$BTC
#AI #数据中心 #ESG

Trade decision: Do not trade, but this is a potential turning-point signal in the AI bubble narrative.

a16z long-form: how solar + batteries are reshaping the grid—and what Base Power is doing.

Key takeaways: There are no affluent countries with low electricity usage. AI data centers are power black holes; traditional energy sources can’t keep up, so solar + storage is the only way out.

AI’s ceiling isn’t compute—it’s electricity.

$BTC
#AI #能源 #infrastructure

Trading view: don’t trade directly, but the AI energy narrative is worth watching long term.

Trump has been reported to frequently call the Federal Reserve Chair, Powell, sometimes several times within a few days. The topics include the Iran war and the impact of AI on the economy.

"Federal Reserve independence" can be taken back now. The president directly calls the Fed chair—this isn’t independence; it’s reporting.

The direction of monetary policy has been hijacked by politics, which is a long-term positive for hard assets.

$BTC
#宏观 #美联储 #policy

Trade outlook: The direction of monetary policy has been hijacked by politics, which is a long-term positive for hard assets like BTC.

Arkham API now adds support for the x402 standard. AI agents can pay for Arkham API usage fees directly with USDC—pay as you go, with no subscription required.

x402 is becoming the payment standard for the Agent economy. From Browser Use to Arkham, more and more APIs are integrating x402, and economic activity among agents is forming a closed loop.

Agents are no longer just tools—they are economic entities.

$USDC
#AI #x402 #Arkham

Transaction assessment: AgentFi infrastructure is accelerating—watch, but don’t chase.

August mass leadership changes? Four major personnel shifts on the same day:

1. Google DeepMind CEO Demis Hassabis will no longer handle day-to-day management, shifting to Chairman + Alphabet’s Chief Scientist
2. Legendary engineer Jeff Dean leaves Google, taking four key people to found Discovery Loop, using AI to fully automate the entire research workflow from start to finish. Alphabet’s share price once fell by more than 5%
3. X product lead Nikita Bier steps down and becomes a consultant
4. Move language’s founder/Sui CTO Sam Blackshear leaves Mysten Labs and joins Anthropic to work on defensive security research

Jeff Dean’s departure is the biggest signal—Google’s AI talent moat is starting to loosen. Sam Blackshear going to Anthropic shows that AI safety is attracting top talent.

$SUI
#AI #人事变动 #Sui

Trade outlook: Losing the CTO is a negative for Sui, but it does not affect the long-term outlook. The AI safety space is worth watching.