Josh TW (@josh_tw) header
Josh TW (@josh_tw) profile picture

Hacking Africa's energy abundance. Cactus Is The New Lithium.

Earth
Joined March 2012
Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)Photo by Josh TW (@josh_tw)
A Hidden Bubble Carries Our Cosmic Home Our Sun carries its planets through the Milky Way inside a protective bubble. Far away, two enormous Fermi Bubbles reveal clues to our galaxy’s turbulent past
25
215
12
1,086
42,843
I had Grok Bot make a chart to visualize it
2
86
11,996
Elon’s technical fluency at a micro-detail level is unparalleled among big-tech CEOs, I think. It’s among his greatest sources of credibility.
2
2
60
6,786
卧槽!看完这个你就知道为什么他们要撸掉老黄了! 老黄被 CSIS 主席 John Hamre 直接当面贴脸开大:“你前阵子说中国正在赢得 AI 竞赛,我们真要输了吗?” 老黄先是笑了,吐槽媒体做标题党总是漏掉前提和依据。 接着他没打太极,直接抛出一个极其硬核的“五层蛋糕”框架,把中美 AI 家底从下往上扒了个精光: 1. 能源(最底层):中国的能源总量,直接是美国的两倍。 2. 芯片:美国确实领先几代。但老黄警告别自满,半导体本质是制造业,千万别以为中国造不出来。 3. 基础设施:美国建个数据中心从破土到上线大概三年,而中国一个周末就能盖座医院。 4. 模型:美国的前沿模型大概领先六个月,但在开源模型上,中国已经遥遥领先。 5. 应用(最顶层):谁先把技术用起来、铺得最广,谁就能赢下这场工业革命。 老黄这个拆解太毒辣了。外行看 AI 竞赛只盯着最上面的聊天机器人,老黄看的却是整条工业栈——拼到最后,底座其实是电力和基建狂魔的速度。就算芯片和前沿模型卡住半年几年的身位,一旦被开源生态和暴力基建追平,应用层的全面开花是谁也拦不住的。 说白了:基建干不过中国,美国建数据中心要三年,中国一个周末盖座医院; 模型也没领先多少,美国前沿领先半年,但中国开源遥遥领先; 最狠的是底层能源——中国的能源总量,是美国的两倍! 看完老黄这五层蛋糕的盘点,你觉得美国靠“芯片+闭源前沿模型”这两层的优势,守得住这场工业革命的胜局吗?评论区聊聊。 (乱局之下,更该重温去年12月3日CSIS这场炉边谈话。)
152
927
83
3,108
226,440
Beautiful Saturn 😍🪐
152
1,334
73
7,591
138,093
Josh TW retweeted
“NVIDIA has to become a frontier lab faster than frontier labs are becoming chip companies.”
Ok I do have a view on NVIDIA without Jensen. I’ll tell you the real reason $NVDA will likely fail without him and why the frontier labs and hyperscalers are the biggest winners. (Elon has been in a really good mood recently. Have you noticed?) The biggest threat to NVIDIA is its own customers, the frontier labs. NVIDIA has to become a frontier lab faster than frontier labs are becoming chip companies. And with Jensen out, they likely won't maintain the momentum to get there. With recursive self-improvement (RSI), every frontier lab and hyperscaler must become a chip company or lose the AI race. It's that simple, and they've all laid the groundwork (Google TPUs, OpenAI and Meta building custom ASICs with Broadcom, Anthropic building an in-house silicon team, and SpaceX/xAI/Tesla building Terafab). That's because software-only RSI has clear physical limits. Software can reduce how much computation and data movement a model needs, but it can't change the chip's physical bandwidth, memory capacity, or interconnect. Custom silicon gives labs another way to improve performance: designing the memory system, compute, and networking around their own workloads. To get the real compounding effect of RSI, you have to co-design the model and the silicon so it becomes a two-way feedback loop: smarter models design better custom chips, and those chips train and run the next model faster. You can't do that on off-the-shelf NVIDIA GPUs because you get zero proprietary edge buying the same chip as everyone else, and getting your next-gen architecture into NVIDIA's silicon means sharing your roadmap with them years before tape-out (which they then sell to your rivals). Plus, AI now automates writing compilers and porting CUDA kernels to other chips, which kills the CUDA moat. Jensen argued custom ASICs aren't more cost-efficient than GPUs because models change too fast for fixed hardware, and partners like Broadcom charge fat margins too. But hardware doesn't fall behind model changes when AI co-designs the next model and chip together instead of humans designing them separately (which Google and OpenAI are already doing). And labs aren't building custom silicon just to save on NVIDIA's 75%+ gross margins. Closing that hardware-software loop is the main goal, though the cost savings are still massive since RSI requires 24/7 internal inference where custom ASICs are way cheaper and more power-efficient than general-purpose GPUs. That's why NVIDIA has been trying so hard to become a frontier open-weight lab, buying Groq (fast inference IP), Poolside + Nemotron (internal coding models), and Hugging Face (distribution). If closed labs win, enterprises just call their APIs running on the labs' own custom silicon and NVIDIA gets cut out. Subsidizing free open-weight models is the only way to make enterprises buy or rent NVIDIA GPUs to run the models themselves. With Jensen gone, it's unclear if the Hugging Face deal even survives DOJ approval (can a GPU company own the whole stack?), and a non-founder is unlikely to take the massive risks needed to become a frontier lab: putting billions of capex into internal clusters and free open-weight models instead of just selling GPUs, while competing directly with its own biggest customers $GOOG, $META, $SPCX, $ANT, $MSFT
236
583
57
9,073
587,007
Boring Company President Steve Davis says Elon Musk knows the material and price of every part on every Tesla car and SpaceX rocket, and calls him "the best engineer in the world" “It’s great. I mean, arguably the best engineer in the world. So, why would you not want to work for the best engineer in the world? I can go anecdote after anecdote after anecdote.” “But I think the one thing that is very visible with Elon is that he knows, as you’ve seen, every minute detail of every single part on every Tesla car, on every SpaceX rocket, and he knows its purpose. And if there is no purpose, it gets deleted.” “So, he’ll see a part and say, what’s that made of? Oh, it’s A514 steel, 100 ksi yield strength, 50 cents a pound. He’ll know all the numbers and he’ll know why it’s there.” “And that’s something that’s really important for engineers to do. So, everyone at the Boring Company can walk up and down the boring machine. They know the material of every part, the yield strength of every part, the cost of the material, the cost of the part, the lead time for the part, the supply chain for the part, and that allows you to make good decisions.” “That sounds super obvious, but I would challenge anyone, even at the company you’re working at. Do you know that for all of the parts? And it’s rare the answer is yes. And for Elon, the answer is unequivocally yes.”
Boring Company President Steve Davis reveals the Boring Company steers its machines with PlayStation controllers instead of $35,000 custom joysticks, from a control room full of gamers "A typical boring machine company will have these $30,000, $35,000 custom joysticks, and if a button breaks, you have to pay like $2,000 and wait six weeks." "So we just use PlayStation controllers, which are super cheap, like $100 if that, and they're replaceable. But almost as important is everyone knows how to use them." "So we live out of our control room, which is a bunch of people playing video games, and they're very, very good at it."
26
232
18
1,695
164,096
🇨🇳 Chinese engineers are developing motorcycle helmets with airbags that inflate the moment a crash begins! 🤯 The company Lanxi Yema is testing sensors that detect impacts and trigger an inflator to cushion the rider’s head and neck. The technology is still being tested, with similar prototypes developed in Europe. Most people who ride every day would take the extra cushion if it actually shows up when they need it. Writer: Lucas
114
188
22
933
142,694
世界一贅沢なクレープ、料理人も凄いけどキャビアとイクラが山盛り
386
1,367
284
8,363
1,115,578
Josh TW retweeted
Advanced solid state transformers at scale from Tesla
Introducing Powercore Solid state medium voltage transformer, inverter and breaker – reliable, safe and secure 7.4 MW/9 MVA continuously at 34.5 kVac w/ grid-forming support for 10MVA of peak power Made in America. Operating 2027
1,160
2,157
178
21,470
2,129,572
SpaceX has caught a Starship coming back from orbit. On November 14, 2026, the tower's arms caught Ship 42 at Starbase. No ship had ever been caught before. How the catch works in five steps, with the real pictures from SpaceX's broadcast, and what Elon Musk says a catch is worth. Plans are shown as plans. Clips: Elon Musk at SpaceX's Third Quarter 2026 Webcast (Oct 20, 2026), the Baron Investment Conference (Nov 6, 2026), the All-In Summit (Sep 15, 2026 and Sep 10, 2025) and X Takeover 2026 (Oct 10, 2026); Gwynne Shotwell at the All-In Summit (Sep 15, 2026); SpaceX's own videos and its Flight 15 broadcast (Nov 13 and 14, 2026). Our 3D is illustrative. Through SpaceX, Elon Musk is compressing global travel to under an hour while building the transportation system to make life multiplanetary. By mastering full rocket reusability and propellant production on Mars, he’s solving the cost barrier of reaching—and returning from—other worlds. Time and again, he has delivered what legacy aerospace said couldn't be done. Welcome to the multiplanetary era. Follow, like and share for the next film. via Now Illustrated #Starship #SpaceX #ElonMusk #Starbase
91
764
48
5,131
310,948
Josh TW retweeted
Antonio Gracias has invested with Elon since PayPal. He sat on Tesla's board through the Model 3 crisis. He says one night explains Elon better than anything. It's 2018, 3 AM, in the body shop. By Elon's own account, Tesla was about a month from bankruptcy. Elon sits down with the engineers and starts cutting welds out of the car. About 30. Math in his head. "If he had been wrong, we'd have gone bankrupt." The simulation came back. He was right. "It was like watching Michael Jordan play basketball times 10." Gavin Baker, sitting next to him, on why Elon started xAI: so AI isn't "just dominated by OpenAI, Google." That panel was two years ago. xAI was a startup with no product.Gavin Baker explained why Elon started xAI before most people took it seriously. "To make sure that it is not just dominated by OpenAI, Google." Then Antonio Gracias, who sat on Tesla's board, explained why Elon can actually pull it off. It's 2018, 3 AM. The Model 3 body shop is failing. Tesla is running out of money. "He was running the math in his head… take about 30 welds out of the car." "If he had been wrong, we'd have gone bankrupt." He wasn't wrong. "It was like watching Michael Jordan play basketball times 10." That's the same brain now running xAI. Back then it was 30 welds. Now it's a million GPUs. Everyone asks if xAI can catch OpenAI and Google. The people who've seen Elon at 3 AM don't ask. Today it's inside SpaceX, running one of the largest GPU clusters on Earth. Both of them called it early. Both had seen 3 AM.
35
323
20
3,058
395,709
"الانذار الاحمر" َ
152
515
9
8,583
1,034,300
身家万亿的老板,正趴在一百米高空,摸一艘刚从太空掉下来的“废铁”。 评论区都在操心:刚落地的火箭,凑这么近不怕炸吗? 其实老马不傻。飞船下午3点20挂上筷子臂,他在塔底下老老实实等了两个半小时,直到安全插销锁死、剩油全部抽干,5点半才坐电梯上去。 真要是去摆拍发朋友圈,落地那一秒就该冲上去了。他非得亲自爬上去,原因很简单: 1. 去看真伤情:这是人类第一次接住从太空轨道回来的飞船。在天上飞了22.5小时,扔了26颗星链V3,再顶着几千度高温冲回地面。隔热瓦掉了几块?被烧成彩虹色的不锈钢外壳有没有内伤?要修多久才能接着飞?这些答案电脑算不出来,PPT上也没有,全写在飞船的伤疤上。 2. 去救半场残局:这次试飞,底下那节助推器因为没油,掉进墨西哥湾炸了。等于今天只成功了一半,这艘飞船是全场唯一的活口。作为总工程师,他不亲手摸一遍底,接下来的发射排期根本没法拍板。 当然,除了去验货,他还有点私心。 23年前,马斯克在斯坦福吹了个牛,说想往火星送一颗种子。 23年后的今天,那艘真正能把种子送去火星的船,第一次从太空活着回来了。 所以他上塔之后,伸手抱了抱这艘刚从火海里捞出来、掉着黑渣的飞船。 换作是你,苦熬23年看到它活着回来,你也得上去抱一下。 你们觉得,星舰离像坐飞机一样当天往返,最难啃的骨头还剩什么?
Inspecting ship 42 right now
278
1,120
61
12,343
1,472,610
Josh TW retweeted
Four months after we launched Manus, I was invited to a dinner with President Macron. I ended up seated directly across from Jensen Huang. He was half an hour late, and Macron asked him to give a toast. Jensen picked up his wine glass and turned it into a fully improvised speech — in front of a table of European politicians and founders. It was electric. The charisma of a top founder, on full display. I typed the whole thing into my Notes that night. Still one of the sharpest takes I’ve heard on software, AI and Europe. What a pity. What a loss.
140
530
61
7,426
584,652
Well damn 🔥
Let’s take this gentleman’s argument at face value that anyone can throw money at rockets and get something like Starship, Falcon or Dragon. Perhaps he can explain to everyone how SpaceX is able to catch giant starships from orbit with chopsticks and deliver 85% of all Earth payload to orbit with Falcon, while Blue Origin, which started before SpaceX and was funded by the immensely wealthy Jeff Bezos, is not yet able to launch reliably? Also, why is it that Boeing and Lockheed received several times what SpaceX did from the government for development of their rockets and spacecraft, but the Boeing capsule is still not operational and ULA delivers less than 5% of SpaceX payload to orbit? I am genuinely curious about these questions.
407
1,425
70
18,217
747,212
Josh TW retweeted
I have a very wide distribution of training data
Former NASA astronaut Garrett Reisman, who spent seven years working with Elon Musk at SpaceX, explains what makes @elonmusk such an extraordinary engineer: “I mean, I’ve met a lot of super, super smart people, but they’re usually super, super smart on one thing.” “And he’s able to have conversations with our top engineers about the software and, you know, the most arcane aspects of that. And then he’ll turn to our manufacturing engineers and have discussions about some really esoteric welding process for some crazy alloy.” “And his ability to do that across all the different technologies that go into rockets and cars and everything else he does. That’s what really impresses me.” “He’s really the most driven person I think I’ve ever met.”
1,010
1,798
92
21,957
2,364,647
One of the coolest things I've ever seen!! Loved the thunderclap as the starship dropped out of the heavens, before the catch. True alien tech. Huge congrats @elonmusk and the @SpaceX team! the whole stack is now proven out. Ad Astra! 🚀
243
747
72
12,613
1,656,203
Josh TW retweeted
Replying to @marsrepublica
The road ahead is long and we are still in the first mile, although it may seem otherwise
442
353
52
6,170
347,875
SpaceX/Tesla just posted this very specific job opening
Bloomberg’s reporting Nvidia’s board is currently meeting and considering replacing CEO Jensen Huang following reports that he had several affairs, including one with a woman suspected of spying for China. Citing sources, Bloomberg reporting The board has hired an outside firm to look for possible successors and is also discussing whether to appoint a temporary CEO while finding a permanent replacement.
254
498
217
7,262
1,189,034