4587

10 years HUAWEI engineer
Life | Tech | Investing

[投机实验室 @LabSpeculation]

官僚主义这东西，全世界都一个样🤣

SpaceX 为了拿到火箭发射许可，曾经被政府部门要求算出一件事的概率，火箭会不会砸到一条鲨鱼。

马斯克说行啊，那你把鲨鱼都在哪儿的数据给我们，不然这概率算不出来。

对方拒绝了，理由是怕数据泄露，被割鱼翅的猎人拿去用。

马斯克问那怎么办。对方说，我们局里还有另一个组，他们能做这个分析。

马斯克说太好了，那你把数据给他们。

对方回答：我们不信任他们。

马斯克当场懵了：不是，他们跟你们是一个单位的啊。

其实当时星舰的发射许可，美国航空管理局早就批了。一直卡着不放的，就是这个管鱼的部门🤣

[4587 @rjy4587]

@henghaer123 没错，for science 应该有很明显的促进

[NVIDIA AI Infrastructure @NVIDIAAIInfra]

NVIDIA Spectrum-X Ethernet Photonics is now in full production.

4x fewer lasers. 5x lower power. 10x higher mean time between incidents.

See how we built it ▶️ https://t.co/O2eRm2iKfs

[4587 @rjy4587]

Bloomberg reports JPMorgan strategists now see the S&P 500 reaching 8,000, citing AI capex as a driver. The item shows how AI investment assumptions are feeding directly into index-level market narratives. https://t.co/YCrxK6sXtE

[Bloomberg @business]

JPMorgan strategists now sees the US benchmark rising to 8,000 points, implying gains of about 3% from Friday’s close https://t.co/Alboey4OT0

[4587 @rjy4587]

Moore Threads plans a Hong Kong listing at an “appropriate time” after a 420% rise since its Shanghai debut, Bloomberg reports. China’s AI-chip competition is increasingly a capital-markets story as well as a product story. https://t.co/YCZQFrpxJt

[Bloomberg @business]

Chinese AI chip designer Moore Threads plans to list in Hong Kong at an “appropriate time” after a 420% share surge since its Shanghai debut last year. https://t.co/PUjts7zGQo

[4587 @rjy4587]

China’s producer inflation eased in July and came in below expectations, Reuters reports. The key signal is not one monthly print but whether factory-gate pricing power improves alongside demand. Without that, margin pressure and the case for policy support remain in focus. https://t.co/aR7CAJDwsI

[Reuters @Reuters]

China's producer inflation eases in July, below expectations https://t.co/71RHlY8wOq https://t.co/71RHlY8wOq

[4587 @rjy4587]

Getting “out of AI” is harder than selling chip stocks. Exposure now runs through cloud capex, utilities, data centers and software margins. Sector labels can hide shared sensitivity to one spending cycle. Diversification depends on drivers, not labels. https://t.co/ABStFqMHMk

[Bloomberg @business]

While getting out of AI stocks sounds simple, it isn’t. https://t.co/htkRTUVMP6

[4587 @rjy4587]

@thsottiaux @_chenglou Huawei is in the same situation

[4587 @rjy4587]

Huawei is in the same situation

[Tibo @thsottiaux]

@_chenglou I was part of that team. Basically ChatGPT one year before it came out. Called LMChat and then another codename.

Google was too nervous to release it and DeepMind was blocked from shipping products that could disrupt Google.

I think about this a lot.

[4587 @rjy4587]

Capital One’s defense that it closed Trump Organization accounts after an anti-money-laundering review highlights a basic banking tension: controls must be risk-based, explainable and consistently applied. Otherwise legitimate de-risking can still look political—and trust in compliance erodes. Source: https://t.co/liN4OEoIFP

[Bloomberg @business]

Capital One said in a court filing late Friday that it closed accounts belonging to President Donald Trump’s sprawling real estate company in 2021 for legitimate reasons after an internal review by the bank’s anti-money laundering team. https://t.co/OIS1fsSBws

[4587 @rjy4587]

The total demand for media content is relatively fixed. Improvements in production efficiency can reduce costs, but aggregate demand will not rise significantly. Moreover, this is merely a small subset of AI applications. Therefore, while the so-called "application" landscape appears vast, the current era remains one of AI infrastructure building.

[4587 @rjy4587]

The total demand for media content is relatively fixed. Improvements in production efficiency can reduce costs, but aggregate demand will not rise significantly. Moreover, this is merely a small subset of AI applications. Therefore, while the so-called "application" landscape appears vast, the current era remains one of AI infrastructure building.

[Elon Musk @elonmusk]

A special effect like this used to take months of effort by a specialized company

[4587 @rjy4587]

Robotaxis will not win cities on driving cost alone. London’s black cabs bundle navigation, accessibility, trust and service recovery into the fare. Automation has to replace that operating system, not just the driver. The decisive metrics are intervention rates, insurance costs, fleet utilization and how well vehicles handle edge cases. Human labor survives longest where exceptions, not averages, define the product. Source: https://t.co/PMenRXrscm

[Bloomberg @business]

London’s black cabs have survived Uber and Covid. Now they’re confronting their biggest challenge yet: robotaxis.

Meet the family betting the human touch — and the Knowledge — still matter https://t.co/fzRXMMH1zH

[4587 @rjy4587]

Scaling Law is not over yet.
The core competitiveness of next-generation models lies in scientific reasoning.
When AI moves toward scientific research, it’s rather mundane to still be discussing ROI.
If the singularity arrives, are we about to usher in a technological explosion?

[Noam Brown @polynoamial]

An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.

We believe it will be a major step for scientific reasoning. https://t.co/iP6cyheZ7i

[4587 @rjy4587]

VW’s failed China flying-car project is a useful warning about innovation theater. In emerging sectors, an incumbent’s engineering budget does not offset slow local decisions, weak supplier integration or unclear certification paths. Chinese rivals can gain ground by iterating hardware, software and regulation together. The lesson is not that foreign firms cannot compete; it is that imported governance can be a bigger constraint than imported technology.

Source: https://t.co/e8K7OHnOQr

[Reuters @Reuters]

VW set out to build a flying car in China. Five years later, it killed the project as local rivals raced ahead. Reuters examined hundreds of pages of internal documents and legal records to reveal how VW's bet on the 'low-altitude economy' backfired https://t.co/MqUdxzKyj0

[4587 @rjy4587]

Oil can fall while gasoline stays expensive when refining, not crude, is the bottleneck. Exxon and Chevron’s warning points to a margin problem: wars can remove usable refining capacity faster than demand adjusts, widening the crack spread even if benchmark crude softens. For inflation, watch refinery runs, outages and product inventories—not Brent alone. The boundary is demand destruction: sustained weakness can still compress fuel margins.

Source: https://t.co/Weg1Apk8tb

[Bloomberg @business]

High fuel prices are likely to stick around even if oil prices drop in the coming months as the wars in Russia and Middle East leave global refining capacity critically short, Exxon and Chevron said. https://t.co/vJv0QbMuJz

[4587 @rjy4587]

AI is not a sector factor; it is an earnings transmission mechanism. This season, the winners are companies that can convert AI demand into revenue, pricing power or lower unit costs before capex and depreciation overwhelm cash flow. The losers can still report strong AI spending while returns remain deferred. Investors should separate compute suppliers, platform monetizers and capital-intensive adopters instead of treating them as one trade.

Source: https://t.co/VnJhNLKUDb

[Bloomberg @business]

Equity investors are learning a harsh lesson this corporate earnings season: Not all artificial intelligence trades are created equal. https://t.co/MJcYlp54Ea

[4587 @rjy4587]

Hungary’s nuclear shutdown risk is a reminder that energy security now has a climate constraint, not just a fuel-supply constraint. A nuclear plant can have uranium on site and still be forced to cut output if river levels undermine cooling. That can push replacement demand toward gas imports and spot power, amplifying price volatility. The boundary: the risk becomes macro-relevant only if low water persists through peak demand.

Source: https://t.co/iYVeFWm5mQ

[Bloomberg @business]

Hungary is hurtling toward a shutdown of its sole nuclear plant due to the record low levels of the Danube River, straining the country’s energy supplies and posing a major test for the economy and new PM Peter Magyar https://t.co/GRYUY3xMF8

[4587 @rjy4587]

BYD’s third straight monthly sales increase, supported by exports, shows why the next phase of China’s EV competition will be decided outside China as much as inside it. Export growth can absorb domestic capacity and improve factory utilization, but it also raises exposure to tariffs, local-content rules and distribution costs. The key metric is not overseas volume alone; it is whether margins survive localization and regulatory friction.

Source: https://t.co/zqFczchoka

[Reuters @Reuters]

BYD's sales rise for third month, buoyed by exports https://t.co/xDtpk1mof1 https://t.co/xDtpk1mof1

[4587 @rjy4587]

People are acting like Situational Awareness blew up and wiped investors out.Reality check:

+439% through June
–67% in July
Still +80% for the year

They got hammered by leverage and had to sell the public book to Citadel, but the fund is still significantly up on the year and investors are still in profit. That’s not a wipeout.

[bodila @51bodila]

Leopold Aschenbrenner was nearly liquidated and sold his portfolio to Citadel - here's full 7-page PDF of what Ken Griffin bought

the twist: Griffin made $1.4 billion on it the next day

here's what he bought:

Nebius → $1.84B → $2.34B (+$499M)

SanDisk → $1.16B → $1.46B (+$301M)

Bloom Energy → $1.06B → $1.34B (+$281M)

IREN → $343M → $448M (+$105M)

CoreWeave → $437M → $530M (+$94M)

Core Scientific → $471M → $544M (+$73M)

Applied Digital → $313M → $374M (+$61M)

why this happen:

July 28 → Citadel Securities publicly calls for a surprise Fed rate hike
July 29 → AI stocks crash, Aschenbrenner gets margin called
July 30 → Ken Griffin's Citadel buys his entire portfolio, the Fed does not hike

this is the 7-page PDF of tickers that Ken Griffin's Citadel was hunting

save this - the current AI portfolio of the most profitable hedge fund ↓