Arthur Welson

Never be shame of where you start, and never set limitations on where you end.

[Mad Animals @sixu495472]

Came for a chill sit, ended up flipping over.#dog

[Hao HONG 洪灝, CFA @HAOHONG_CFA]

With the National Day long holiday just around the corner, the Shanghai Composite and the Hang Seng are trying hard to take back the lost ground at the open today. In the near term, 3800 is a dependable bottom for traders.

Consensus seems to think yesterday’s stimulus was underwhelming and calls for “structural reform” to solve the problem. However, the structural problem of strong investment and weak consumption has existed for years. Using a long-term structural problem to explain and resolve short-term issues is illogical and futile.

I have long held that weak consumption is an income and wealth distribution and disparity problem, as Chinese labor compensation as a percentage of output is way too low. I have consistently expressed this view in various high-level speeches and interviews in the past few years. Simply put, Chinese labor is underpaid, especially considering how productive it has been.

To reform is to improve. It's good that the Chinese economy has grown dramatically, and thus can now afford to pay its labor more. The money made from exports should be used better for people, their welfare, healthcare, and development. $FXI $KWEB $CQQQ $BABA $PGJ $PDD $EWH

[Outer Space Today @outerspacetoday]

JUST IN: The U.S. Space Force is building new software designed to detect satellite maneuvers almost immediately, rather than waiting hours for perfectly refined orbital tracking data.

Mission Delta 2 says the system will continuously fuse data from military and commercial radars, optical telescopes and radio-frequency sensors to flag unexpected orbital changes, close approaches and rendezvous activity across LEO, MEO and GEO.

The shift is deliberate: commanders would get faster, probabilistic warnings even if the first measurements are less precise, then see the track automatically tighten as more sensor data arrives.

That matters as satellites become more maneuverable and capable of rapid proximity operations. The Space Force’s own 2040 framework says space-domain awareness is moving from a safety-of-flight function toward real-time battlespace awareness and engagement support.

Testing continues through late 2026, with integration into operational centers planned for 2027.

[Mario Nawfal @MarioNawfal]

🇰🇷 South Korea just built a welding robot that doesn’t need scaffolding.

This is the Diden Spider, a four-legged machine with magnetic feet that can walk across steel floors, climb vertical walls, and stick to ceilings.

It carries up to 30 kg, squeezes through tight shipyard spaces, and welds where people used to need ladders, harnesses, and a lot of risk.

Shipyards in Korea are already putting it to work.

One more tough, dangerous job that machines are starting to take over.

Writer: Daniyal

[National Geographic @NatGeo]

New cat alert! A call from a wildlife sanctuary about a "weird cat" has led to the first new cat species discovered in over 100 years in Bolivia’s Yungas forest ecoregion. Learn more about this feline surprise: https://t.co/crffXvPqH4

[Teslaconomics @Teslaconomics]

I just had my Grok Bot pull together the best clips of Starship sitting in the ocean after recovery.

And I’ve watched this part of @SpaceX's “Holy Grail of Rocketry” film over and over.

Bc when you isolate these shots, the scale of what SpaceX is doing really hits you.

A full Starship sitting in the middle of the ocean.

SpaceX employees standing on it and next to it looking absolutely tiny.

Bro... that ship just went to space.

And it came screaming back through Earth’s atmosphere faster than a bullet.

And it survived reentry.

And it splashed down beautifully in the ocean.

And then humans went out there to recover this giant stainless-steel spaceship, looking almost like an alien spaceship that just landed.

You see, for most of the history of rocketry, rockets were basically one-shot machines.

We would launch them and then lose them.

SpaceX is fighting tooth and nail to completely change that with a goal to build a rocket system where the booster comes back, the ship comes back, you turn them around, and you fly them again. And again. And again.

That is the "holy grail" of space exploration.

We’re watching reusable spaceflight slowly turn from something that sounds impossible into something people will one day look at as completely normal.

SpaceX is making this a reality... and this is the stuff that inspires me so much.

[momoai沫沫🫧 @momoai_daily]

我已抵达翻译软件最深处

LinkedIn speak真是个神奇的小语种

我将全职在家研究✋😭🤚

[Phoenix Yin @Phoenixyin13]

为什么前沿模型放出的成果几乎都是清一色的数学证明，物理定律的发现与突破等于0？

到现在为止，人类物理领域最后一次框架级事件是 1925–1926年的矩阵力学/波动力学。
如果宽松一点，我可以算到 1973年的渐近自由和标准模型闭合。之后的50年时间，物理学家数量涨了两个数量级，经费涨了三个，产出的框架级结果是 0。

如果人类以百倍的智力投入、半个世纪，也没做出来，那这件事的困难就不在认知端。
指望AI模型做到，等于指望它解决一个难度完全位于认知之外的问题。

数学的真理判据在问题内部；
物理的真理判据在世界外部。

数学特别适合当前 AI，是因为它近乎形成了一个完美的封闭环境。
给你 axioms、definitions、conjecture，然后模型搜索证明，最后一步甚至可以被 Lean、Isabelle、Coq的形式验证器机械判定。
于是，AI 可以做海量搜索、失败、反馈、再搜索。Reward 极其干净：proof checked = 1，proof failed = 0。
这恰好是强化学习、搜索、test-time compute 最喜欢的问题结构。

物理完全不是这样。
AI模型无论进化多少代，架构与训练网络不变，它就永远写不出牛顿三大定律和万有引力定律，我可以确信以及肯定这件事。
因为现代的LLM 极度偏向 interpolation，而基本物理革命往往要求 representation break。前者的训练目标要求它学习人类已有分布，所以产生一个真正脱离已有 ontology 的 representation，比在既有 representation 内做推理难得多。
即使它能写出类似的东西或定理，你无法在把它丢给一个“Physics Verifier”之后，机器告诉你 TRUE。
目前看来，一个基本物理假说可能要等 CERN、LIGO、JWST、量子实验室或者新的精密仪器给它一个数据点。
这造成了一个极其严重的feedback bandwidth mismatch。

假设今天把整个物理学文献、实验数据库、arXiv 都给 GPT-级模型，它会看到成千上万个 anomaly：
3 sigma deviation、材料中的反常行为、Hubble tension、muon anomaly、暗物质候选、超导异常、核结构问题……

最后的结果是，AI 越完美地学习现有物理学家的 priors，它越可能成为一个极其优秀的正常科学家。
但是，正常，意味着它永远无法做出范式突破。
目前的模型，无论Astra也好，还是没有彻底公开的Mythos，仍然生活在text、code、images和datasets里。
但是，自然科学家生活在world。
Faraday 可以不停摆弄线圈和磁铁，Rutherford 可以看散射数据，Michelson 可以重新设计光学装置。

实验科学中，有一个非常强大的能力叫intervention，拥有这种能力会每天问自己：如果我主动改变世界，会发生什么？
AI，没有操纵现实、改造实验的欲求。

这就是我们今天看到的趋势，AI 数学能力会先于 AI 基础物理发现能力十几年成熟。
Google DeepMind 的 AlphaEvolve 已经开始把能力扩展到模拟和科学问题，但其最漂亮的成果仍然集中在那些具有明确 objective function、可以高速计算验证的领域。
但距离物理定律级别的发现，AI还有太长的路。

[Ammanichanda @Arkasiraee]

There is a reason enormous steam and gas turbines still play an important role in modern power grids, even as renewable power generation grows.

A 1,000+ MW class steam turbine generator spinning at 3,000 RPM isn't just generating electricity. Its rotating shaft line can store roughly 4-5 giga joules of kinetic energy.

That matters because electricity generation and consumption have to remain continuously balanced.

If demand suddenly rises, power plants cannot instantly produce more steam, burn more fuel or increase water flow to ramp up power generation.

Instead, the electrical load on synchronised generators increases immediately, and hundreds of enormous rotating machines across the grid collectively surrender a fraction of their stored kinetic energy.

Their rotors slow almost imperceptibly as grid frequency begins falling below 50 Hz.

This is synchronous inertia.

It buys the critical first seconds for governors, hydro, batteries, gas turbines and operating reserves to increase output and restore the balance. Without sufficient response, frequency can continue falling until protection systems begin disconnecting loads and generators leading to a power grid failure or widespread blackout.

And this balancing act can span thousands of kilometres, connecting millions of consumers and generators.

Yet electrically, the entire grid still has to behave like one enormous synchronised machine.

[FatParrot @fatparrotdaily]

If they have to, the Chinese will cut through the Himalayas to find the dead.
From Lhasa to Shigatse to Gyirong Port, Highways 219 and 216 on the roof of the world are packed with excavators from provinces across China—about 15 kilometers of machines, a lot of them staged as relief crews waiting to rotate onto the front. Civilian flights keep pouring people and supplies into Lhasa. The mud the machines have already pulled out has been piled into a 200-meter hill behind the line. In 48 hours, mechanized units have dug six kilometers toward the port. They’ve fully uncovered buried Highway 216. One kilometer left to Gyirong.
At the gate itself, hundreds in the advance team are searching around the clock for anyone still alive. Temples across Tibet are praying. Ordinary Chinese people have donated on their own.
A handful of Indians have smeared China and Nepal online. That is not the weight or the values the Ganges civilization is supposed to stand for. Politics is a tumor in human history.
No force on earth can run a dig-and-rescue like this at Gyirong, in the Himalayas. If anyone can—it’s the Chinese people.

[Arthur Welson @Arthur_Welson]

地形决定了千年古道不会因泥石流而抛弃，而是用更科学的基建方案减少损失，比如峡谷内只建道路 仅通货物 无人车驾驶 流量管控 修建坡上避难所等...

[骆逸 @royxy]

@pietrosa4 只有避开灾害区才是唯一的办法。这种冰川崩塌灾害非常罕见，但只要有概率，灾害就一定会重现。估计口岸要是重建就不会盖在这里了

[China Xinhua News @XHNews]

Rescue teams have reached the hardest-hit area in Gyirong, southwest China's Xizang.
Firefighters used ropes, boats and drones to navigate the severely damaged terrain after a devastating mudslide.
Search and rescue efforts are ongoing. #Xizang #China #Rescue #DisasterResponse

[박선달 | Park Sundal @parksundal0528]

빨간색으로 표시된 지역의 위엄

전 세계 인구의 20%
전 세계 제조업 생산량의 40%
아시아 전체 수출의 54%

[Massimo @Rainmaker1973]

Before and after of Syapru Beshi, one of the affected areas of today's flash flooding in Nepal.

[DJI @DJIGlobal]

Never seen a rocket launch up close? Experience the awe with Osmo 360 II — featuring a full 1-inch omnidirectional sensor, native 8K resolution, and 2.4 μm large pixels for unprecedented clarity and impact. ⚠️ Professional operation. Do not attempt.

[Jason Z @JasonZX]

机器人参加一次运动会的代价也太大了，跑完，就终结者了。

[LandSpace @LandSpace_Tech]

ZhuQue-3 Y2 Launch Mission Recap

- The Return of ZhuQue | 朱雀归来

[LandSpace @LandSpace_Tech]

MISSION SUCCESS | ZhuQue-3 Y2 Reusable Launch Vehicle Achieved Full Success in Orbital Insertion and First-Stage Recovery

On August 19, 2026, at 07:35 (UTC+8), the ZhuQue-3 (ZQ-3) Y2 reusable launch vehicle lifted off from the Dongfeng Commercial Space Innovation Pilot Zone. Approximately 137 seconds after lift-off, the first and second stages separated. The second stage continued its flight and successfully delivered the Honghu 03 satellite, independently developed by Hongqing Technology, into its designated orbit. At approximately 07:41, the first stage performed a successful soft touchdown at the LandSpace Landing Site#1 in Minqin County, Gansu Province, following the planned trajectory — marking the full success of the flight test mission.

[Tz @Tz_2022]

这个 NASA 发布的大洋环流动态图简直就是梵高星空的再世。。。