Binance Square
Sleepy
93 Posts

Sleepy

Square Verified
Editor-in-Chief https://beating.news Covering AI, youth culture, and how emerging tech is changing the way young people live.
3 Following
149 Followers
432 Liked
Posts
·
--
The so-called AI CapEx bubble—at least at OpenAI, reality might be just the opposite: last year, they spent less money. OpenAI CFO Sarah Friar recently said that the compute power they acquired a year ago would be worth 3 to 5 times as much if they were to sell it on the market today. But OpenAI won’t sell, because what they lack most right now is compute power. Friar even said they should have bought more at the time. For years, the market has questioned the logic behind AI giants pouring in tens of billions of dollars into data centers and GPUs: with how fast the technology iterates, won’t the compute power bought today turn into a pile of depreciating assets tomorrow? At least the results seen so far are nothing like that. The compute that OpenAI locked in early last year didn’t rapidly depreciate—instead, due to tight supply it generated a scarcity premium of 3 to 5 times. At the same time, the company still lacks compute power itself.
The so-called AI CapEx bubble—at least at OpenAI, reality might be just the opposite: last year, they spent less money.

OpenAI CFO Sarah Friar recently said that the compute power they acquired a year ago would be worth 3 to 5 times as much if they were to sell it on the market today.

But OpenAI won’t sell, because what they lack most right now is compute power. Friar even said they should have bought more at the time.

For years, the market has questioned the logic behind AI giants pouring in tens of billions of dollars into data centers and GPUs: with how fast the technology iterates, won’t the compute power bought today turn into a pile of depreciating assets tomorrow?

At least the results seen so far are nothing like that.

The compute that OpenAI locked in early last year didn’t rapidly depreciate—instead, due to tight supply it generated a scarcity premium of 3 to 5 times. At the same time, the company still lacks compute power itself.
·
--
Article
SpaceXAI is planning to turn the release event for its AI Agent into a real-life startup reality show.SpaceXAI is planning to turn the release event for its AI Agent into a real-life startup reality show. From September 15 to 17, Matt Palmer, Lauren Tan, and Roshan Sadanani will continuously livestream for three days, using Grok Bot to “build a company” from a blank page. Not to prepare the product in advance and then go on stage to demo. The process officially given is: first think about what to build, then write a business plan, define the functions, truly write the code—take it all the way to the final product showcase. The event page simply says: Can three people build a company in three days? This is actually more interesting than running another benchmark.

SpaceXAI is planning to turn the release event for its AI Agent into a real-life startup reality show.

SpaceXAI is planning to turn the release event for its AI Agent into a real-life startup reality show.
From September 15 to 17, Matt Palmer, Lauren Tan, and Roshan Sadanani will continuously livestream for three days, using Grok Bot to “build a company” from a blank page.
Not to prepare the product in advance and then go on stage to demo. The process officially given is: first think about what to build, then write a business plan, define the functions, truly write the code—take it all the way to the final product showcase. The event page simply says:
Can three people build a company in three days?
This is actually more interesting than running another benchmark.
·
--
Verified
Article
Suiyuan Technology went public today. The most worth watching figure for this stock isn’t the USD 9.1 billion valuation—it’s the 4.16% free float.Suiyuan Technology went public today. The most worth watching figure for this stock isn’t the USD 9.1 billion valuation—it’s the 4.16% free float. This Tencent-affiliated AI chip company raised RMB 6.119 billion in its IPO—about USD 912 million. The issue price was RMB 142.18 per share, corresponding to a listed market cap of roughly RMB 61.2 billion, or about USD 9.1 billion. After issuance, the company has a total share capital of 430 million shares; today, only 17.9 million shares are actually available for trading on the market. The float ratio is 4.1595%. This makes Suiyuan’s stock price quite interesting today. The market seems to be pricing an AI chip company worth more than RMB 60 billion, but in the short term, the number of shares available for trading is only a small fraction—about one twenty-something of the total share capital. Add in the current hype around domestically made AI chips, and the price can easily be amplified by this tiny group of tradable shares.

Suiyuan Technology went public today. The most worth watching figure for this stock isn’t the USD 9.1 billion valuation—it’s the 4.16% free float.

Suiyuan Technology went public today. The most worth watching figure for this stock isn’t the USD 9.1 billion valuation—it’s the 4.16% free float.
This Tencent-affiliated AI chip company raised RMB 6.119 billion in its IPO—about USD 912 million. The issue price was RMB 142.18 per share, corresponding to a listed market cap of roughly RMB 61.2 billion, or about USD 9.1 billion.
After issuance, the company has a total share capital of 430 million shares; today, only 17.9 million shares are actually available for trading on the market.
The float ratio is 4.1595%.
This makes Suiyuan’s stock price quite interesting today.
The market seems to be pricing an AI chip company worth more than RMB 60 billion, but in the short term, the number of shares available for trading is only a small fraction—about one twenty-something of the total share capital. Add in the current hype around domestically made AI chips, and the price can easily be amplified by this tiny group of tradable shares.
·
--
Article
The wildly popular overseas personal agent tool, Instinct, is raising more funding again.The wildly popular overseas personal agent tool, Instinct, is raising more funding again. On August 26, it just closed a $250 million round, valuing the company at $2.5 billion. It’s only been about two weeks. According to the latest report from The Information, the founder, Noah Shinn, has already been discussing raising another $1 billion with investors, with a valuation that could potentially reach $10 billion. Two weeks, valuation at 4x. And Instinct still hasn’t been fully released to the public yet—it still mainly relies on invite codes. The reason behind this round of funding is also quite simple: there isn’t enough compute capacity. An Instinct personal agent isn’t really that different from a typical chatbot. If you ask it to book a restaurant, buy something, handle emails, or cancel a subscription, it will actually go operate the computer and phone. Behind a single user request, it may run a long chain of agent actions—so as usage grows, compute consumption will surge as well.

The wildly popular overseas personal agent tool, Instinct, is raising more funding again.

The wildly popular overseas personal agent tool, Instinct, is raising more funding again.
On August 26, it just closed a $250 million round, valuing the company at $2.5 billion. It’s only been about two weeks. According to the latest report from The Information, the founder, Noah Shinn, has already been discussing raising another $1 billion with investors, with a valuation that could potentially reach $10 billion.
Two weeks, valuation at 4x.
And Instinct still hasn’t been fully released to the public yet—it still mainly relies on invite codes.
The reason behind this round of funding is also quite simple: there isn’t enough compute capacity.
An Instinct personal agent isn’t really that different from a typical chatbot. If you ask it to book a restaurant, buy something, handle emails, or cancel a subscription, it will actually go operate the computer and phone. Behind a single user request, it may run a long chain of agent actions—so as usage grows, compute consumption will surge as well.
·
--
Article
OpenAI releases ChatGPT for Financial Services.OpenAI releases ChatGPT for Financial Services. It’s a ChatGPT Work customized for the financial industry, designed with the participation of Morgan Stanley and Evercore, with the first batch targeting investment banking and equity research teams directly. The most critical layer is actually the data. Daloopa, PitchBook, LSEG News, and Crunchbase data are built in directly, and OpenAI is responsible for indexing and hosting; existing subscriptions such as S&P Capital IQ, LSEG, MSCI, Factiva, and Moody’s are also connected under a unified login. The entire system also has 50+ connectors, so you can continue pulling data from companies such as Datasite, Preqin, Box, and Intapp.

OpenAI releases ChatGPT for Financial Services.

OpenAI releases ChatGPT for Financial Services. It’s a ChatGPT Work customized for the financial industry, designed with the participation of Morgan Stanley and Evercore, with the first batch targeting investment banking and equity research teams directly.
The most critical layer is actually the data.
Daloopa, PitchBook, LSEG News, and Crunchbase data are built in directly, and OpenAI is responsible for indexing and hosting; existing subscriptions such as S&P Capital IQ, LSEG, MSCI, Factiva, and Moody’s are also connected under a unified login. The entire system also has 50+ connectors, so you can continue pulling data from companies such as Datasite, Preqin, Box, and Intapp.
·
--
Article
One thing about Anthropic that really annoys me is that it truly believes that as long as it thinks it’s doing good, it can make many decisions for users.One thing about Anthropic that really annoys me is that it truly believes that as long as it thinks it’s doing good, it can make many decisions for users. It just released a 154-page Threat Intelligence Report, compiling cases where Claude has been used for cyberattacks, scams, biological research, weapons development, model distillation, and more. How did these cases come about? Anthropic looks at how users use Claude. After it identifies accounts that it believes are problematic, it continues investigating—analyzing how these people use it—then determines whether the user has crossed the line. If it thinks there’s risk, it suspends the account. It then compiles these events into a report, explaining to the outside world what it found and what it prevented.

One thing about Anthropic that really annoys me is that it truly believes that as long as it thinks it’s doing good, it can make many decisions for users.

One thing about Anthropic that really annoys me is that it truly believes that as long as it thinks it’s doing good, it can make many decisions for users.
It just released a 154-page Threat Intelligence Report, compiling cases where Claude has been used for cyberattacks, scams, biological research, weapons development, model distillation, and more.
How did these cases come about?
Anthropic looks at how users use Claude. After it identifies accounts that it believes are problematic, it continues investigating—analyzing how these people use it—then determines whether the user has crossed the line. If it thinks there’s risk, it suspends the account. It then compiles these events into a report, explaining to the outside world what it found and what it prevented.
·
--
Article
Seven domestic AI chip companies that are listed or preparing to be listed generated about 21.5 billion yuan in revenue in the first half of the year. Among the six companies that disclosed their inventories, there are already 23.9 billion yuan worth of inventory sitting on their books.Seven domestic AI chip companies that are listed or preparing to be listed generated about 21.5 billion yuan in revenue in the first half of the year. Among the six companies that disclosed their inventories, there are already 23.9 billion yuan worth of inventory sitting on their books. Five companies that disclosed operating cash flow burned a total of about 6.4 billion yuan in cash in the first half of the year, and only Cambricon is still positive. This 23.9 billion yuan can’t simply be understood as “chips can’t be sold.” HuaWei? Actually Cambricon is a good example. Of its inventory with an original value of about 9 billion yuan, 5.7 billion is still raw materials, 2.5 billion is in the contract-manufacturing stage, and the truly finished goods and goods already shipped and awaiting acceptance are only about 700 million yuan.

Seven domestic AI chip companies that are listed or preparing to be listed generated about 21.5 billion yuan in revenue in the first half of the year. Among the six companies that disclosed their inventories, there are already 23.9 billion yuan worth of inventory sitting on their books.

Seven domestic AI chip companies that are listed or preparing to be listed generated about 21.5 billion yuan in revenue in the first half of the year. Among the six companies that disclosed their inventories, there are already 23.9 billion yuan worth of inventory sitting on their books.
Five companies that disclosed operating cash flow burned a total of about 6.4 billion yuan in cash in the first half of the year, and only Cambricon is still positive.
This 23.9 billion yuan can’t simply be understood as “chips can’t be sold.”
HuaWei? Actually Cambricon is a good example. Of its inventory with an original value of about 9 billion yuan, 5.7 billion is still raw materials, 2.5 billion is in the contract-manufacturing stage, and the truly finished goods and goods already shipped and awaiting acceptance are only about 700 million yuan.
·
--
Proud of the ADHD era for us
Proud of the ADHD era for us
·
--
Article
What Palantir truly wants to do now is to distill decades of a company’s experience directly into a model.What Palantir truly wants to do now is to distill decades of a company’s experience directly into a model. Today, NVIDIA and Palantir have taken another step forward—Nemotron has been integrated into Foundry/AIP, and the first pilot customer is NVIDIA itself. The scenario is still the hardest-to-tackle part of the supply chain. This scenario is very complex—one Vera Rubin rack alone has about 1.3 million parts, backed by thousands of suppliers. If one board, a cooling component, or a batch of memory is missing, delivery of the entire machine could be delayed. Palantir first organizes data on orders, inventory, suppliers, and production capacity, then uses NVIDIA’s own supply-chain data to further train Nemotron. Going forward, the model won’t just learn supply-chain knowledge—it will also learn how NVIDIA manages inventory, identifies bottlenecks, and weighs delivery against cost.

What Palantir truly wants to do now is to distill decades of a company’s experience directly into a model.

What Palantir truly wants to do now is to distill decades of a company’s experience directly into a model.
Today, NVIDIA and Palantir have taken another step forward—Nemotron has been integrated into Foundry/AIP, and the first pilot customer is NVIDIA itself. The scenario is still the hardest-to-tackle part of the supply chain.
This scenario is very complex—one Vera Rubin rack alone has about 1.3 million parts, backed by thousands of suppliers. If one board, a cooling component, or a batch of memory is missing, delivery of the entire machine could be delayed.
Palantir first organizes data on orders, inventory, suppliers, and production capacity, then uses NVIDIA’s own supply-chain data to further train Nemotron. Going forward, the model won’t just learn supply-chain knowledge—it will also learn how NVIDIA manages inventory, identifies bottlenecks, and weighs delivery against cost.
·
--
Article
Paul Christiano joins the OpenAI Foundation board of directors, also enters Safet, and will also join the OpenAI Group PBC board of directors as a non-voting observer.Paul Christiano joins the OpenAI Foundation board of directors, also enters Safet, and will also join the OpenAI Group PBC board of directors as a non-voting observer. His assessment is already extremely aggressive: based on recent capability progress and the difficulty of Alignment, AI capabilities can quickly accelerate in the short term, and ultimately there will be a "catastrophic, irreversible loss of control"—a risk that is no longer something that can be ignored. His subjective probabilities for the next year are 4%, and for the next three years are 15%. The inflection point Paul is most worried about is AI R&D automation. If AI starts doing AI research on its own, algorithmic progress will produce stronger and more AI researchers, who then, in turn, continue improving the algorithms, forming a capability feedback loop.

Paul Christiano joins the OpenAI Foundation board of directors, also enters Safet, and will also join the OpenAI Group PBC board of directors as a non-voting observer.

Paul Christiano joins the OpenAI Foundation board of directors, also enters Safet, and will also join the OpenAI Group PBC board of directors as a non-voting observer.
His assessment is already extremely aggressive: based on recent capability progress and the difficulty of Alignment, AI capabilities can quickly accelerate in the short term, and ultimately there will be a "catastrophic, irreversible loss of control"—a risk that is no longer something that can be ignored.
His subjective probabilities for the next year are 4%, and for the next three years are 15%.
The inflection point Paul is most worried about is AI R&D automation.
If AI starts doing AI research on its own, algorithmic progress will produce stronger and more AI researchers, who then, in turn, continue improving the algorithms, forming a capability feedback loop.
·
--
Article
This leaderboard for V4.1 Flash actually reveals DeepSeek’s next most important route shift:This leaderboard for V4.1 Flash actually reveals DeepSeek’s next most important route shift: Large models don’t necessarily have to be that smart, but they must be better at getting things done. Look at the most traditional intelligence benchmarks—V4.1 Flash has even regressed. GPQA is 90.9, below V4 Pro’s 92.4; bare HLE is 36.8, while V4 Pro is 42.7. Compared with GPT-5.6 Sol and Claude Opus 5, the gap is even more obvious. But as soon as you put the model into a computer and let it start working, the curves suddenly flip. Terminal-Bench 4.0 goes from 12.4 → 31.2, DeepSWE from 62.7 → 74.2, Automation-Bench from 43.2 → 54.8, Agents' Last Exam from 25.7 → 31.8, and ExploitGym from 5.4 → 15.3. In almost every project that reflects an agent’s execution ability, V4.1 Flash has pulled away from last month’s V4 Pro by a wide margin.

This leaderboard for V4.1 Flash actually reveals DeepSeek’s next most important route shift:

This leaderboard for V4.1 Flash actually reveals DeepSeek’s next most important route shift:
Large models don’t necessarily have to be that smart, but they must be better at getting things done.
Look at the most traditional intelligence benchmarks—V4.1 Flash has even regressed. GPQA is 90.9, below V4 Pro’s 92.4; bare HLE is 36.8, while V4 Pro is 42.7. Compared with GPT-5.6 Sol and Claude Opus 5, the gap is even more obvious.
But as soon as you put the model into a computer and let it start working, the curves suddenly flip.
Terminal-Bench 4.0 goes from 12.4 → 31.2, DeepSWE from 62.7 → 74.2, Automation-Bench from 43.2 → 54.8, Agents' Last Exam from 25.7 → 31.8, and ExploitGym from 5.4 → 15.3. In almost every project that reflects an agent’s execution ability, V4.1 Flash has pulled away from last month’s V4 Pro by a wide margin.
·
--
Looks like we don't need to get a new phone.
Looks like we don't need to get a new phone.
·
--
Article
If Tomorrow Human Civilization Were to Be ExtinctIf tomorrow human civilization were to be extinct If at four o’clock tomorrow afternoon, humanity will disappear, I would probably first think of the wind of September in Beijing. If at four o’clock tomorrow afternoon, humanity will disappear, I would probably first think of the wind of September in Beijing. It slips through the gaps between buildings; it’s no longer as scorching as high summer. In the morning, people heading out begin to hesitate about whether to wear a light outer layer. The corner of the plastic awning over the street-side breakfast stall gets lifted by the wind. The leaves are still green, and sunlight glints on the windshield of a car. At noon, someone complains that the delivery for their takeout has been slow. In the subway, someone keeps their head down, playing games. A cat lies on the windowsill, basking in the sun. It doesn’t know that humanity will have no tomorrow.

If Tomorrow Human Civilization Were to Be Extinct

If tomorrow human civilization were to be extinct
If at four o’clock tomorrow afternoon, humanity will disappear, I would probably first think of the wind of September in Beijing.
If at four o’clock tomorrow afternoon, humanity will disappear, I would probably first think of the wind of September in Beijing.
It slips through the gaps between buildings; it’s no longer as scorching as high summer. In the morning, people heading out begin to hesitate about whether to wear a light outer layer. The corner of the plastic awning over the street-side breakfast stall gets lifted by the wind. The leaves are still green, and sunlight glints on the windshield of a car. At noon, someone complains that the delivery for their takeout has been slow. In the subway, someone keeps their head down, playing games. A cat lies on the windowsill, basking in the sun. It doesn’t know that humanity will have no tomorrow.
·
--
Article
With this iPhone Duo from Apple today, the biggest takeaway I got was that the design team makes me want to buy it like crazy, while the product manager keeps desperately trying to talk me out of it.With this iPhone Duo from Apple today, the biggest takeaway I got was that the design team makes me want to buy it like crazy, while the product manager keeps desperately trying to talk me out of it. First, let me praise it a bit. The moment you unfold the screen is genuinely really cool. Apple’s UI extension from the outer display to the inner display, along with the transitions and opening/closing animations, are so smooth. In that instant when the 7.6-inch screen opens up, I haven’t felt this kind of strong desire to buy a phone just from one action alone in a long time. After all these years of foldable phones, this is the first time I’ve felt that the act of “folding” itself can be an experience—not just a way to force in a bigger screen.

With this iPhone Duo from Apple today, the biggest takeaway I got was that the design team makes me want to buy it like crazy, while the product manager keeps desperately trying to talk me out of it.

With this iPhone Duo from Apple today, the biggest takeaway I got was that the design team makes me want to buy it like crazy, while the product manager keeps desperately trying to talk me out of it.
First, let me praise it a bit.
The moment you unfold the screen is genuinely really cool. Apple’s UI extension from the outer display to the inner display, along with the transitions and opening/closing animations, are so smooth. In that instant when the 7.6-inch screen opens up, I haven’t felt this kind of strong desire to buy a phone just from one action alone in a long time.
After all these years of foldable phones, this is the first time I’ve felt that the act of “folding” itself can be an experience—not just a way to force in a bigger screen.
·
--
Article
NVIDIA has recently quietly rewritten a very important narrative:NVIDIA has recently quietly rewritten a very important narrative: GPUs may not even be valued according to the logic of traditional consumer electronics. NVIDIA AI Infrastructure today described GPUs with three words: productive, fungible, durable. Then added one more line: an investable asset class. One of the biggest debates around the AI bubble over the past two years has been: how long can the GPUs—bought for tens of billions of dollars—actually be used? NVIDIA releases new cards every year. If older cards become economically obsolete in just two or three years, then the return on the entire AI CapEx would look very bad.

NVIDIA has recently quietly rewritten a very important narrative:

NVIDIA has recently quietly rewritten a very important narrative:
GPUs may not even be valued according to the logic of traditional consumer electronics.
NVIDIA AI Infrastructure today described GPUs with three words:
productive, fungible, durable.
Then added one more line: an investable asset class.
One of the biggest debates around the AI bubble over the past two years has been: how long can the GPUs—bought for tens of billions of dollars—actually be used? NVIDIA releases new cards every year. If older cards become economically obsolete in just two or three years, then the return on the entire AI CapEx would look very bad.
·
--
Article
What happens after Anthropic calculates what AI would do when it pushes the U.S. economy to its limits.Anthropic has just calculated what would happen after AI pushes the U.S. economy to its limits. The conclusion is a bit bleak: AI can create an economic miracle and, at the same time, drive a decline for workers. In Anthropic's most aggressive scenario, by 2030 the United States’ GDP is 32.4% higher than the baseline without “AI.” This growth rate is already close to an economic takeoff. But workers’ total income is only 0.5% higher than the baseline. Where did the money go? Capital income is 81.4% higher, and capital’s share of national income rises from about 40% to 54.8%. In other words, most of the new wealth created by AI is taken by capital.

What happens after Anthropic calculates what AI would do when it pushes the U.S. economy to its limits.

Anthropic has just calculated what would happen after AI pushes the U.S. economy to its limits.
The conclusion is a bit bleak:
AI can create an economic miracle and, at the same time, drive a decline for workers.
In Anthropic's most aggressive scenario, by 2030 the United States’ GDP is 32.4% higher than the baseline without “AI.” This growth rate is already close to an economic takeoff.
But workers’ total income is only 0.5% higher than the baseline.
Where did the money go?
Capital income is 81.4% higher, and capital’s share of national income rises from about 40% to 54.8%. In other words, most of the new wealth created by AI is taken by capital.
·
--
Verified
Uniswap has started to go after the money of arbitrage bots. On September 10, Uniswap Labs launched the StablePair Hook on Ethereum mainnet. The first pairs were USDC/USDG and USDC/USDT. In Q2, Uniswap’s stablecoin swaps totaled $43.4 billion—more than all the next three on-chain trading venues combined. Previously, stablecoin pools had a very consistent loophole. After the USDC/USDT price deviated from 1:1, arbitrage bots would come in and push the price back. In the meantime, most of that price gap went to the bots, while LPs only earned a single fixed fee. StablePair Hook now directly targets this money. When the price is within the normal range, it dynamically adjusts fees to maintain a fixed bid-ask spread. When the price runs out of the range, any trade that pushes the price further outward is charged 0 fees. Arbitrage that truly pulls the price back starts an auction. At first, the fee rate is set high, then decreases block by block, until an arbitrage bot thinks there’s enough profit to act. Bots can still earn, but they no longer have to take all the correction profits for themselves—what remains is left for LPs. In reality, this is already very close to the logic of on-chain market makers: the fee rate isn’t a fixed number—it changes with the price position, and even “who will correct the price, and how much profit the correction can generate” is dynamically priced. By this point, Uniswap v4 Hooks are getting more and more interesting. In the past, AMMs only had curves and fees; now, even how arbitrage profits are split can be written directly into the pool’s rules.
Uniswap has started to go after the money of arbitrage bots.

On September 10, Uniswap Labs launched the StablePair Hook on Ethereum mainnet. The first pairs were USDC/USDG and USDC/USDT. In Q2, Uniswap’s stablecoin swaps totaled $43.4 billion—more than all the next three on-chain trading venues combined.

Previously, stablecoin pools had a very consistent loophole. After the USDC/USDT price deviated from 1:1, arbitrage bots would come in and push the price back. In the meantime, most of that price gap went to the bots, while LPs only earned a single fixed fee.

StablePair Hook now directly targets this money.

When the price is within the normal range, it dynamically adjusts fees to maintain a fixed bid-ask spread. When the price runs out of the range, any trade that pushes the price further outward is charged 0 fees. Arbitrage that truly pulls the price back starts an auction.

At first, the fee rate is set high, then decreases block by block, until an arbitrage bot thinks there’s enough profit to act. Bots can still earn, but they no longer have to take all the correction profits for themselves—what remains is left for LPs.

In reality, this is already very close to the logic of on-chain market makers: the fee rate isn’t a fixed number—it changes with the price position, and even “who will correct the price, and how much profit the correction can generate” is dynamically priced.

By this point, Uniswap v4 Hooks are getting more and more interesting. In the past, AMMs only had curves and fees; now, even how arbitrage profits are split can be written directly into the pool’s rules.
·
--
MOO surged suddenly yesterday—up more than 80% roughly every 24 hours. Its market cap is now about $42 million, with trading volume around $6.4 million. What it corresponds to is Micron MU, and the name is very straightforward too: MU sounds like “moo,” and when you pair it with the hottest storage-chip行情 (market trend) lately, it’s turned directly into “Memory Cow Moo.” Over the past few days, on the Robinhood Chain, these “US stock + meme” coins have already been pumped through a round—MOO with MU, Artificial Inu with NVDA, BONER with HIMS. The gameplay isn’t new. What allowed MOO to break out this time was mainly that Micron and the storage sector line itself has been hot. And its MOO/MU pool has already grown quite sizable. In the past 30 days, out of about $107 million of on-chain trading volume on Robinhood Chain involving MU, MOO/MU contributed about $31.4 million—nearly 30%. Now the meme-and-stock gameplay is already pretty simple: whichever US stock is hot, go find the matching meme next to it. These past two days, it’s MU’s turn.
MOO surged suddenly yesterday—up more than 80% roughly every 24 hours. Its market cap is now about $42 million, with trading volume around $6.4 million.

What it corresponds to is Micron MU, and the name is very straightforward too: MU sounds like “moo,” and when you pair it with the hottest storage-chip行情 (market trend) lately, it’s turned directly into “Memory Cow Moo.”

Over the past few days, on the Robinhood Chain, these “US stock + meme” coins have already been pumped through a round—MOO with MU, Artificial Inu with NVDA, BONER with HIMS. The gameplay isn’t new. What allowed MOO to break out this time was mainly that Micron and the storage sector line itself has been hot.

And its MOO/MU pool has already grown quite sizable. In the past 30 days, out of about $107 million of on-chain trading volume on Robinhood Chain involving MU, MOO/MU contributed about $31.4 million—nearly 30%.

Now the meme-and-stock gameplay is already pretty simple: whichever US stock is hot, go find the matching meme next to it.

These past two days, it’s MU’s turn.
·
--
CLARITY has been changed again, and this time there’s a line that’s definitely worth a serious DeFi project taking a close look at. A newly leaked draft from the latest Republican proposal in the U.S. Senate is set to clarify something that used to be rather vague: talking about being DeFi isn’t enough—you have to look and see whether anyone in the background can actually be in charge. If a protocol still has a team—or a group of people—behind it that can modify features, change operating rules, or has the power to restrict or review users’ usage, then it may be deemed “non-decentralized DeFi,” requiring registration with the CFTC. Going forward, whether a project counts as DeFi may no longer depend on how it describes itself on its website, whether it has a DAO, or whether it has issued governance tokens. Instead, it will hinge on several very specific questions: Who holds the upgrade authority? Who can change the rules? Who can keep users out? Many protocols that today call themselves DeFi can’t survive questions like these. The new draft also makes two additional changes: the DeFi-related provisions have been narrowed to spot and cash transactions involving digital commodities, mainly to address disputes arising from prediction markets; at the same time, it also adds permissions for credit unions to participate in business related to digital assets. According to Lummis, this draft has already included more than 100 amendments proposed by Democrats. But that doesn’t mean the bill is secured. On September 15, the Senate will first hold a procedural vote, requiring 60 votes. The Republicans have 53 seats—so even if they all support it, they would still need at least 7 votes from Democrats. So what’s truly worth watching right now isn’t whether someone else is out there shouting that CLARITY must pass. The bigger shift has already been written into the text: the U.S. is trying to turn “decentralization” from a term defined by an industry itself into a set of standards that regulators can use to check item by item. Once this is truly written into law, its impact on DeFi will likely last much longer than a few points’ worth of up-or-down movement on the day of a vote.
CLARITY has been changed again, and this time there’s a line that’s definitely worth a serious DeFi project taking a close look at.

A newly leaked draft from the latest Republican proposal in the U.S. Senate is set to clarify something that used to be rather vague: talking about being DeFi isn’t enough—you have to look and see whether anyone in the background can actually be in charge.

If a protocol still has a team—or a group of people—behind it that can modify features, change operating rules, or has the power to restrict or review users’ usage, then it may be deemed “non-decentralized DeFi,” requiring registration with the CFTC.

Going forward, whether a project counts as DeFi may no longer depend on how it describes itself on its website, whether it has a DAO, or whether it has issued governance tokens. Instead, it will hinge on several very specific questions: Who holds the upgrade authority? Who can change the rules? Who can keep users out?

Many protocols that today call themselves DeFi can’t survive questions like these.

The new draft also makes two additional changes: the DeFi-related provisions have been narrowed to spot and cash transactions involving digital commodities, mainly to address disputes arising from prediction markets; at the same time, it also adds permissions for credit unions to participate in business related to digital assets.

According to Lummis, this draft has already included more than 100 amendments proposed by Democrats. But that doesn’t mean the bill is secured.

On September 15, the Senate will first hold a procedural vote, requiring 60 votes. The Republicans have 53 seats—so even if they all support it, they would still need at least 7 votes from Democrats.

So what’s truly worth watching right now isn’t whether someone else is out there shouting that CLARITY must pass.

The bigger shift has already been written into the text: the U.S. is trying to turn “decentralization” from a term defined by an industry itself into a set of standards that regulators can use to check item by item.

Once this is truly written into law, its impact on DeFi will likely last much longer than a few points’ worth of up-or-down movement on the day of a vote.
·
--
Tonight’s CPI could directly determine whether the U.S. Federal Reserve will raise rates next week. Beijing time 20:30, the U.S. will release August CPI—also the last major inflation data set before the FOMC on September 15–16. Market consensus from a Reuters survey is: CPI month-over-month +0.4%, year-over-year +3.4%; Core CPI month-over-month +0.2%, year-over-year +2.4%. What really matters is the core monthly rate of 0.2%. Waller recently made his position very clear: if inflation runs “too hot,” he would consider supporting a September rate hike. Reuters’ view is that if tonight’s core CPI comes in at the expected 0.2%, it would likely be enough for the Fed to hold steady. After yesterday’s PPI release, market pricing for a 25bp hike next week has already been pushed to roughly 70%. What’s more troublesome now is oil. WTI and Brent both surged above $100 yesterday: WTI closed at $102.48 and Brent at $107.63. Of course, this newly unfolding spike in oil prices won’t be directly factored into tonight’s August CPI. But the market will be watching for one thing: have the energy price increases already seen in August started to feed through and lift prices for other goods and services as well? If the core monthly rate comes in at 0.2%, there may still be room for rate-hike expectations—which rose after last night’s PPI—to come down a bit. But if it’s 0.3% or even 0.4%, the probability of a rate hike next week will keep moving higher. BTC is still hovering around the $78,000 range. If tonight’s CPI is hotter than expected, it will be even harder for BTC to get back above $80,000. If it’s cooler, then last night’s selloff triggered by the PPI release would have a chance to catch its breath.
Tonight’s CPI could directly determine whether the U.S. Federal Reserve will raise rates next week.

Beijing time 20:30, the U.S. will release August CPI—also the last major inflation data set before the FOMC on September 15–16. Market consensus from a Reuters survey is:

CPI month-over-month +0.4%, year-over-year +3.4%;
Core CPI month-over-month +0.2%, year-over-year +2.4%.

What really matters is the core monthly rate of 0.2%.

Waller recently made his position very clear: if inflation runs “too hot,” he would consider supporting a September rate hike. Reuters’ view is that if tonight’s core CPI comes in at the expected 0.2%, it would likely be enough for the Fed to hold steady.

After yesterday’s PPI release, market pricing for a 25bp hike next week has already been pushed to roughly 70%.

What’s more troublesome now is oil.

WTI and Brent both surged above $100 yesterday: WTI closed at $102.48 and Brent at $107.63. Of course, this newly unfolding spike in oil prices won’t be directly factored into tonight’s August CPI. But the market will be watching for one thing: have the energy price increases already seen in August started to feed through and lift prices for other goods and services as well?

If the core monthly rate comes in at 0.2%, there may still be room for rate-hike expectations—which rose after last night’s PPI—to come down a bit. But if it’s 0.3% or even 0.4%, the probability of a rate hike next week will keep moving higher.

BTC is still hovering around the $78,000 range. If tonight’s CPI is hotter than expected, it will be even harder for BTC to get back above $80,000. If it’s cooler, then last night’s selloff triggered by the PPI release would have a chance to catch its breath.
Log in to explore more content
Join global crypto users on Binance Square
⚡️ Get latest and useful information about crypto.
💬 Trusted by the world’s largest crypto exchange.
👍 Discover real insights from verified creators.
Email / Phone number
Sitemap
Cookie Preferences
Platform T&Cs