trying to manufacture inference hw + sw @fdotinc @scale_at_gmi

SF/Taipei
Joined December 2021
I love inference! Tap in🧑🏻‍💻
Welcoming @solidwillity to SCALE! 🚀 His startup, Touchdown Labs, makes AI inference optimization simple and accessible for resource-constrained organizations. Can't wait to see what you build! #SCALEatGMI #GMICloud #Startup
4
19
50,326
William Chen retweeted
Donald Trump calls Jensen Huang “a very handsome guy” and credits him with building the world’s most valuable company “So, he’s still a young guy, handsome guy. Look at him. He’s a very handsome guy." "But as an example, in a fairly short period of time, he’s built, I think, the most valuable—right?—in the stock market. You know, it’s very easy to value." "The most valuable company anywhere in the world." "And it’s basically new technology. It’s not oil and gas and all the other things that you associate with also a lot of money. It’s technology. And this man has done a great job.” _____ Source: DRM News
Sam Altman and Tibo gift everyone a banked reset by pressing the official big blue button “Finally, with all of the new things we shipped today, because we want you all to build, we figured you might need another banked reset.” “So we have actually done so many of these that Tibo has been campaigning to rename us the Reset company. I think he has even made swag, but more importantly, I think he is out here in the audience somewhere. Can we find him? Oh, there.” [Applause as Tibo is found, Tibo stands and shows reset button] Sam: “I think it would be really nice if we could do a low-key reset for everybody here. No confetti, can’t answer anything. Maybe we could do it. Can you do it?Is that a real thing?” Tibo: “What do you think? Should we just press it?” Sam: “Yeah, I think so.” Tibo: “Does it just work like that? We just press it and it happens?” Sam: “I don’t know. Let’s see what happens..." ["Three, two, one.” Reset button pressed] “Thank you for that. That’s obviously for everyone in the world, not just people here at Dev Day.”
5
12
5
281
92,153
William Chen retweeted
We said we planned to rewrite the rules of silicon. We meant it. Today, we’re sharing the first details of our Core First approach. At NUVACORE, we can build a significant portion of the foundational CPU core IP before locking in an instruction set, giving our team the freedom to focus first on building the best core possible. We’ve assembled a team with a track record of building some of the world’s best CPUs. Now we’re starting to pull back the curtain on what they’re building.
10
32
19
275
103,552
its starting
I’ll explain the new Pro 200 plan differently, before I start live tweeting from DevDay on things that are going out! Today we are going to ship a number of things that increase what you can do across the Plus and Pro plans. A lot of compute is online for this increase. As we increase the floor, we are changing the relative difference between plans to be Plus = 1X Pro 100 = 5X Pro 200 = 10X and we are reopening subscriptions for Pro 200 (we had paused it). If you have an existing plan you will keep the 20X multiplier for a bit and also receive a lot of additional credits because we know changes are hard even if it means that everyone will get more in the end.
118
William Chen retweeted
🦔Someone is buying Japanese used books by the ton. Not novels or comics. Philosophy, history, medical law, Edo-period cultural texts. Multiple buyer accounts all ship to one logistics center in Okayama Prefecture that won't answer questions. Export records show over 50 tons, roughly 100,000 volumes, shipped to the US since last year. No buyer has been identified. Separately, court documents confirmed Anthropic's "Project Panama" bought millions of books in the US, cut the spines off, scanned them, and shredded the originals. An internal memo said the goal was to "destructively scan all the books in the world." Similar operations have been reported across Europe. My Take Anthropic's US book operation came out in court filings earlier this year. Project Panama. Millions of books, cut the spines off, scanned, shredded. Their VP chose the codename so nobody outside the company would find out. That story broke and apparently the same operation just moved to Japan. Anonymous buyer accounts, a warehouse in Okayama that won't take questions, and 100,000 volumes of specialized academic texts that nobody buys to resell. I don't know how you build a trillion-dollar industry on material you had to acquire in secret through middlemen because you knew the public would object. Authors spent years on those books and now they're fed through a scanner and thrown away so a chatbot can sound smarter. If the data was free to use, they wouldn't need codenames and anonymous warehouses. The lawsuits over who owns this material have barely begun and I think the copyright exposure across the whole AI industry is enormous. Anthropic's internal memo said the goal was every book in the world, which describes the foundation of the product. Hedgie🤗
185
1,796
261
4,369
658,884
🇨🇳🇺🇸 #China and the #US Reach Eight Deliverables and Understandings 1⃣The two sides agree to build “a constructive China-U.S. relationship of strategic stability on the basis of respect, fairness, and reciprocity.” 2⃣China and the United States support each other in successfully hosting the APEC Economic Leaders’ Meeting and the G20 Summit. The two heads of state intend to attend each other’s meetings. 3⃣The two heads of state agree that Iran should honor its commitment to not developing nuclear weapons, and no country or institution can be allowed to impose tolls on international waterways. 4⃣The two heads of state recall that China and the U.S. were allies in World War II and fought side by side to win the war. 5⃣The two heads of state recognize the positive role of the China-U.S. economic and trade consultation mechanism, endorse the outcomes of the consultations between the two economic and trade teams, including establishing and taking forward the Board of Trade and other mechanisms, reaching an arrangement for a US$30 billion reciprocal tariff reduction, and the extension of the outcomes of economic and trade consultations in Kuala Lumpur, and give instructions that they be implemented. 6⃣The two sides note that cooperation between the counter-narcotics and law enforcement agencies of China and the U.S. has produced visible outcomes. Recently, the two sides working closely together have jointly cracked multiple cases involving New Psychoactive Substances (NPS) and precursor chemicals, and apprehended dozens of relevant criminal suspects in the two countries. 7⃣The two sides agree on the establishment of the China-U.S. AI Dialogue to exchange views on risks and benefits related to AI. The next exchange will occur in November 2026. China and the U.S. also agree to establish a bilateral communication channel for AI incidents. 8⃣The United States welcomes the arrival of two giant pandas, leased by China to Zoo Atlanta.
China and the United States Reach Eight Deliverables and Understandings 🇨🇳🇺🇸 fmprc.gov.cn/eng/xw/zyxw/202…
154
498
185
3,294
922,058
William Chen retweeted
once again those nvidia export controls were so dumb exactly as predicted they now have an ai process that has zero dependency on the usa
🚨 DeepSeek V5 Leak: Beats Astra >DeepSeek is reportedly preparing an imminent V5 launch >Founder Liang Wenfeng calls it the company's biggest bet yet >Rumored at 2 trillion parameters (not 3T) >Reportedly the first DeepSeek model to train fully on Huawei Ascend chips instead of Nvidia >Needs roughly 4x more Ascend accelerators than the Nvidia equivalent to hit the same training scale DeepSeek is reportedly keeping the open-weight strategy
229
219
19
6,670
532,745
Fab fab fab fab
I had some fascinating conversations with engineers here in Silicon Valley today. The most interesting topic was Terafab. Based on what someone at a semiconductor equipment company told me, it sounds like things may be starting to move in earnest. I’ll share more on Substack soon! (And I really hope I can offer subscriptions on X soon too!!!!)
1
3
869
what im doing
Lior Susan spent a year in China He left convinced the West needs a dramatically new approach to building physical industries: “China essentially built a ~$20T economy based on the physical world.” “They built the entire system on these 5 forces: talent, policy, capital, technology, & customer demand.” “They align those 5 forces in a way that lets their economical flywheel fly & it's actually very hard to compete against them as a standalone company.” “My view was: if we don't do it in the Western world, we're screwed.”
1
349
William Chen retweeted
The meeting was one of Friendship, Strength, and Success, for both China and the U.S.A.
1,168
2,675
210
14,033
487,627
William Chen retweeted
Legacy media REFUSED to show you a SINGLE SECOND of President Trump's speech and historic state dinner with Xi Jinping tonight at the White House So here's the ENTIRE thing, from start to finish, with translations for Xi's speech DO NOT LET LEGACY MEDIA KEEP YOU IN THE DARK
2,861
35,123
1,110
119,643
9,674,833
William Chen retweeted
China has now released its official press release of today's Xi-Trump meeting at the White House. At the moment, the US side's press release is not out yet. I read the Chinese version as soon as it came out. It is not particularly long, but it is surprisingly dense. And one thing stood out to me: Beijing is not simply talking about "cooperation." It is laying out a framework for how China and the United States can compete and win together. And the structure of Xi's remarks is revealing: Trade: cooperate. AI: compete, but cooperate. Taiwan: red lines. Read together with the written statement Xi issued upon arriving in Washington yesterday, today's readout may offer the clearest explanation yet of what Beijing means by a "constructive relationship of strategic stability." Here is how I read it. As in my previous posts, the italicized text comes from the Chinese official release, the bolded text highlights key points, and everything else is my analysis. -- Xi pointed out that the exchange of visits between the heads of state of China and the United States in less than half a year is unprecedented in the history of China-US relations. President Trump and I have maintained communication through various channels and have jointly been steering the great ship of China-US relationship. In May this year, President Trump and I jointly agreed to work toward building a constructive China-US relationship of strategic stability. The new vision reflects the realities of China-US ties, embodies our aspirations for the future, and represents an important step toward finding the right way for the two countries to get along with each other. We need to work toward a relationship featuring cooperation as the mainstay, competition with bounds, manageable differences, and promises of peace, so as to find the right way for two major countries to coexist. China and the United States should, and fully can, achieve mutual benefit and win-win results through practical cooperation, help each other succeed through healthy competition, properly manage differences by seeking common ground while respecting differences, and build peace by fostering mutual trust, so that a constructive China-US relationship of strategic stability can better benefit the two peoples and people around the world. There is a lot packed into these three paragraphs. First, the reference to the "unprecedented" frequency of leader-level exchanges is notable. Xi's framing moves from his personal communication with Trump, to sustained communication between the two presidents, and then to something much larger: a mechanism for stabilizing the bilateral relationship itself. Then comes the metaphor of the "great ship." This is not new. Xi has used the ship metaphor in previous meetings or phone calls with Trump. A ship this large does not turn quickly. And if it veers badly off course, the consequences can be enormous. I think the stage on which this "ship" is sailing is the broader Pacific. The ship may encounter storms and rough seas, but the two countries' top leaders are presenting themselves as jointly responsible for keeping it on course. And then there is a phrase that appears twice: "finding the right way." I don't think that repetition is accidental. Beijing is trying to define what coexistence between two competing major powers should actually look like. And the next sentence gives us the formula: "cooperation as the mainstay, competition with bounds, manageable differences, and promises of peace" was already included in Xi's statement after arriving in the United States the day before. It provides a concrete framework for managing the competition and building this new type of relationship. In many ways, this is an attempt to answer a question that has long complicated China-US relations: Do China and the United States have to choose between cooperation and confrontation? Beijing's answer is no. And more importantly, it is beginning to spell out what that alternative might look like. -- Xi stressed that the economic and trade teams of the two countries held a new round of consultations and reached a new joint arrangement, which is good news for industries and people of both countries, as well as the global economy. Cooperation is a two-way street. The stability of US-China economic and trade relations requires both sides to expand the list of cooperation, while shortening the list of problems. As long as the two sides uphold the spirit of equality, mutual respect and mutual benefit, a way out of problems can be found. Xi expressed his hope that the two sides will work in the same direction, stay committed to the principle of mutual benefit and win-win cooperation, and promote the sustained, stable and sound development of China-US economic and trade relations. And this is where the broader idea of "strategic stability" starts becoming concrete. How do you prevent competition from turning into confrontation? Beijing's answer begins with two lists: The cooperation list. The problem list. In my view, expanding the cooperation means continually broadening the areas in which both sides can achieve tangible results. Reducing the problem list, meanwhile, means preventing new disputes from continuously piling onto the bilateral relationship. This is very pragmatic to manage the relationship. The goal is not to build a relationship completely free of friction-that is clearly unrealistic. The real goal is this: To keep increasing the number of areas where the two sides can cooperate, while preventing disputes from expanding to the point where they overwhelm the entire bilateral relationship. And that helps explain why Xi repeatedly emphasizes equality, mutual respect, and mutual benefit. In addition, the phrase "cooperation is a two-way street" carries a significant message: Beijing is signaling that it is willing to cooperate, but it does not believe strategic stability can be built by China alone. One side cannot keep extending the cooperation list if the other side keeps extending the problem list. Meaningful progress requires both sides to engage, compromise, and reciprocate. -- Xi said that while China and the United States, as both major powers in AI, compete in AI, they have even more reasons to cooperate. The two sides should leverage respective strengths in AI rather than guarding against each other. We can continue to engage in dialogue on AI, exchange views on its risks and benefits, and work together to prevent the misuse and abuse of AI. The two sides should maintain human control over AI technologies, uphold a people-centered approach and the principle of using AI for good, and ensure that AI serves human progress. What happens when an issue belongs on both lists at the same time? That is where AI becomes particularly interesting. The AI competition between the two countries are obvious. But the risks of AI do not respect national borders. Misuse, loss of human control, safety risks, and broader questions of governance cannot necessarily be contained within China or the United States alone. Two countries can compete fiercely over who develops the most advanced AI systems while still sharing an interest in preventing those systems from creating risks neither side can control. This is why Xi said the two countries have even more reasons to cooperate. And the remark also echoes President Trump's statement on social media that he would discuss "super intelligence" today. Trade is about expanding the cooperation list. AI sits on both lists. But Taiwan is different. Let's move on. -- Xi stressed that the Chinese side's position on safeguarding national reunification and territorial integrity is clear-cut. It is hoped that the US side will adhere to the correct position of opposing "Taiwan independence," and handle the Taiwan question with caution, so as to lay a solid foundation for China-US strategic cooperation and the development of bilateral relations. The Taiwan question finally came into focus. It's different from the previous topics. There had previously been speculation in the United States that Xi's asking Trump to stop arms sales to Taiwan based on the 1982 US-China Communique. I do not know whether that report was accurate. But the language of the communique is quite explicit: "Having in mind the foregoing statements of both sides, the United States Government states that it does not seek to carry out a long-term policy of arms sales to Taiwan, that its arms sales to Taiwan will not exceed, either in qualitative or in quantitative terms, the level of those supplied in recent years since the establishment of diplomatic relations between the United States and China, and that it intends gradually to reduce its sale of arms to Taiwan, leading, over a period of time, to a final resolution. " Clearly, 44 years is long enough to qualify as a long term. Yet US policy has failed to follow through on the promise made in that document. Looking at the structure of Xi's remarks, Taiwan has surely been placed within the boundaries of strategic stability. After outlining substantial areas for cooperation, the Chinese side is also sending Washington a clear message: Strategic stability does not mean stability without red lines. -- Trump said that the state visit of Xi to the United States is a great visit. China and the United States are both major countries, and he and Xi are good friends who maintain in-depth and close communication. The US-China relations will surely be handled well under the leadership of the two heads of state. The latest round of economic and trade consultations between teams of both countries achieved new outcomes, and both sides should continue dialogue to reach a better agreement. Artificial intelligence concerns the future of humanity, and China and the United States should maintain dialogue and strengthen cooperation on this issue. Trump's response, as presented in the Chinese readout, is notable for what it reinforces. He described the visit as "great." This effectively signaled to the outside world that the meeting did not need to be portrayed as a tense, difficult, or deeply divided encounter. Judging from the narrative presented in the Chinese readout, the two sides have established fairly clear areas of common ground on several fronts: - High-level communication should continue. - Economic and trade negotiations should continue. - Dialogue on AI should continue. - The overall stability of bilateral relations needs to be maintained. At the same time, it is equally important to note that Trump's remarks were relatively broad and general, while Xi's remarks played a more substantial role in defining the broader framework of the relationship. -- Both heads of state also exchanged views on major international and regional issues including the Middle East situation, Ukraine crisis and Korean Peninsula. The two leaders also confirmed that the two countries will support each other in hosting the 2026 APEC Economic Leaders' Meeting and the G20 Leaders' Summit. I would not pass over this final section too quickly. The Middle East. Ukraine. The Korean Peninsula. None of these issues is simply about China-US bilateral relations. Their inclusion reminds us that the relationship between Washington and Beijing does not exist in isolation. China and the United States are permanent members of the United Nations Security Council, two of the world's largest economies, and leading technological powers. Whether they can maintain channels of communication therefore has implications far beyond trade, technology, or Taiwan. Both are deeply embedded in a broader multilateral system. The immediate references to the APEC Economic Leaders' Meeting in China's Shenzhen and the G20 Leaders' Summit in America's Miami reinforce that point. They place the bilateral relationship in a wider context: even as China and the United States compete with each other, they must also continue engaging within institutions and forums where their decisions affect countries far beyond their own. How will Beijing and Washington manage the competition in a healthy way? When the cooperation list and the problem list are both on the table, which one will Washington choose to make longer? There is no answer to that question tonight. What comes next will.
President Xi Jinping left Beijing on the afternoon of September 23. After crossing the Pacific, he landed outside Washington, also on the afternoon of September 23. By the clock, it almost seemed as though no time had passed. There is something strangely appropriate about that image. China and the United States today often appear extraordinarily far apart. Yet neither country is ever truly distant from the other. And before Xi had even begun his meetings in Washington, something unusual happened. President Trump, who just returned from NYC on the mid-night, did not wait for Xi at the White House. He went to Joint Base Andrews himself. For a sitting US president, personally greeting a visiting foreign leader at the airfield is a rare gesture rather than a routine matter of protocol. In the context of visits by Chinese leaders to the United States, it was unprecedented in history. And it marked only the second instance since JFK went to this airport in 1962 to receive British Prime Minister Harold Macmillan. The symbolism was difficult to miss. Before the formal talks had even begun, Washington was already signaling that it attached considerable political importance to the visit and wanted to establish a positive atmosphere for the discussions ahead. I also noticed a warm moment at the arrival ceremony: Peng raised her hand to indicate someone's height while speaking with Melania, and the two were smiling. I could not help wondering whether they might have been talking about Barron. Xi, making his 11th visit to the United States in 41 years, issued a written statement upon arrival. To the best of my recollection, this was also the first time he had formally released such a statement in the United States. I have not yet found a full official English translation, so I translated the remarks myself and offer some analysis here. Xi's remarks are italicized; key points are highlighted in bold; all other text reflects my own analysis. --I am very pleased to pay a state visit to the United States of America at the invitation of President Trump. On behalf of the Chinese people, I would like to extend sincere greetings and best wishes to the American people. China and the United States are both great countries, and the two peoples are both great peoples. The two peoples have maintained friendly exchanges for a long time and their interests are deeply intertwined. China and the United States should be partners, not rivals. The great rejuvenation of the Chinese nation and making America great again can go hand in hand. The two countries are not mutually exclusive, can help each other succeed and advance the well-being of the whole world. The history of direct China-US engagement can be traced back to 1784. At the time, with American trade still constrained by Britain, US merchants decided to sail directly to China. Their ship, the Empress of China, eventually arrived in Guangzhou and was warmly received. Chinese and American accounts from the period contain strikingly positive descriptions of the other side. For more details, see the earlier post I shared in the comments. Chinese thought, especially the teachings of Confucius, was valued by many of America's founding generation, including Franklin, Jefferson and Adams. Confucius is even depicted on the pediment of the US Supreme Court, a reminder of the intellectual curiosity with which some early Americans viewed Chinese civilization. Chinese laborers played a crucial role in building the transcontinental railroad across the United States, a project that would have been extraordinarily difficult without them. Since the nineteenth century, American universities have helped educate generations of Chinese students. Tsinghua University, sometimes described as "China's MIT," was established with important American involvement through the Boxer Indemnity scholarship program, while Yenching University, one of the institutions that later contributed to the development of Peking University, dubbed "China's Harvard," was founded with substantial American support. During World War II, China and the United States fought on the same side against fascism. In the later decades of the Cold War, the two countries again found significant strategic common ground. Since China's reform and opening in 1978, China has supplied American consumers with vast quantities of affordable, high-quality goods, while many US companies have achieved remarkable commercial success in China, some on a scale they had never reached in their home market. China, in turn, benefited from access to American technology, capital, management practices, etc. The point is not that the relationship has ever been free of conflict. It has not. The point is that competition and cooperation have coexisted in it almost from the beginning. The phrase "the great rejuvenation of the Chinese nation and MAGA can go hand in hand" is striking because it deliberately adopts Trump's own political language. Beijing is straightforward: China's rise does not have to come at the expense of America's decline, and no need for the United States to contain China. It also carries an implicit message: China is not asking the United States to abandon its own national ambitions; by the same token, Washington should not automatically treat China's pursuit of a better life as a threat. The formulation also leaves room for future deals: If both sides can accept the premise that both countries can succeed, then issues such as trade, investment, technology, energy, and agriculture can, at least in principle, can be treated as matters of bargaining, exchange, and negotiated interests rather than as existential conflicts. The most important sentence in Xi's statement, comes in the second paragraph: "China and the United States should be partners, not rivals." Xi has made this point to the US many times before. What is different this time is that he appears to have offered a more concrete framework for what that partnership could mean. Now look at the third paragraph: --This year is a critical one for the development of both China and the United States. During President Trump's visit to China in May, we agreed to build a constructive China-US relationship of strategic stability, providing strategic guidance for the future development of bilateral ties. China and the United States should work in the same direction, promote a stable relationship in which cooperation remains the mainstay, competition stays within bounds, differences remain manageable, and peace remains within reach. China and the United States, as two major countries, should find a right way to get along in the new era. I am confident that we can achieve this. Notice what Xi is not saying. The statement is not pretending that China-US competition can disappear. It is not asking Washington to abandon competition. The real emphasis is on placing limits around that competition. Competition may continue, but it should have boundaries. Differences may persist, but they should remain controllable. Neither side has to give up its ambitions, but neither should define the other's success as its own defeat. That is the framework Beijing appears to be putting on the table, and clearly, the framework agreed upon at the Beijing meeting in May should now move from principle to further implementation. --The world is developing and times are changing, but the historical logic of peaceful coexistence between China and the United States has not changed, the goodwill of friendly exchanges between the two peoples has not changed, and the general expectations of the international community for China and the United States have not changed. I look forward to having in-depth exchanges with Trump on major issues concerning bilateral relations and the world situation, constantly enriching the connotation of the constructive China-US relationship of strategic stability, and expanding the space for exchanges and cooperation in various fields between the two countries. I also looks forward to engaging extensively with people from all walks of life in the United States to deepen friendship. With joint efforts, I believe the visit is sure to be fruitful, and will enhance the well-being of the two peoples and add hope for world peace. The three things that have not changed underscore a broader point: whatever changes may occur in the international environment, China and the United States should not decouple to the point of outright hostility. Xi's reference to maintaining broad contacts with people from all walks of life in the United States and deepening friendship suggests a willingness to engage not only with federal officials, but also with the business community, academia, and the general public. This is also consistent with how Xi has approached the United States in the past. When he visited as vice president, he made a point of returning to the small town in Iowa he had first visited as a young official and meeting old friends there. In 2015, he presented a medal to an American veteran who bravely flew missions over the Hump during WWII, and over the years he has repeatedly met with members of the American business and academic communities. More recently, he just wrote back to a group of young Americans. If I had to reduce the entire speech to a single sentence, I would put it this way: Beijing is trying to move the China-US relationship away from a framework of strategic confrontation and unlimited competition, and toward one in which competition remains possible but bounded, differences remain real but manageable, the overall relationship remains stable, and cooperation continues to occupy a meaningful place. That, in my view, may be the most important test of this visit: whether competition can remain bounded without coming to define the entire relationship.
85
410
88
1,408
716,003
William Chen retweeted
What struck me about this photo is that Trump and Xi invited to their dinner table: an immigrant from South Africa, two Taiwanese immigrants that came to the US as children, and a gay kid from Alabama, all of whom moved to Silicon Valley to make a dent in the universe.
369
816
125
11,550
1,025,390
William Chen retweeted
Everyone talks about the need for a system approach in AI infrastructure, but in order to do so, one needs a decent mental model of the system itself. The problem is that most system-level knowledge tends to be gatekept behind domain-specific silos and $3,000 conference passes. IC Engineers are actively disincentivized from seeking those out since many companies narrowly define roles and fail to justify conference expenses for them. Without a mental framework, dealing with system-level issues often runs the risk of staying shallow. As a result, most engineers treat system-level issues not as an elaborate, delicate balance of trade-offs, but high-level requirement management. Most of it involves gathering feedback from domain-specific experts over whether certain requirements are feasible or not. My website helps combat technological and career stagnation by giving engineers sufficient system context across the AI infrastructure stack. I broke through these barriers this past year developing these working mental models from industry-leading conferences and hope that people benefit from it.
1
1
13
1,427
William Chen retweeted
$SPCX: Insane Time to Compute Elon is bringing up 220k GB300s per month which is 500MW IT load per month. How does he do it? Well Tesla has been designing electrical distribution and liquid cooling in cars for 20+ years now. That’s on a moving object in a compact space. Datacenters is stationary object and larger space. Supply chain? Megapack is a 44B/yr revenue product line that uses electrical distribution and liquid cooling components that is datacenter grade. Elon can just take from his existing Megapack supply chain. The liquid piping efficiency in Megapack is world class by the way.
Replying to @minchoi
Colossus 1 is 150k H100, 50k H200 and 30k GB200. Colossus 2 is 110k GB200 and 440k GB300. Another 220k GB300 will be fully operational next week and another 220k in November. If we get lucky, yet another 220k GB300 by late December.
13
14
2
166
52,961
William Chen retweeted
Z.ai founder is Tsinghua professor - lectures to packed classroom. Feels the AGI!
智谱创始人唐杰教授清华开课,走廊、地上都坐满了人。 课后作业是从零手搓Transformer,这才是年轻人该追的星🫡
2
40
1
424
43,007
I’m solving this
Yes meta has a terrible reputation in terms of user trust. But it doesn’t matter. Users will still give away all their data to meta. Here is why: Not a single startup can afford to make personal agent AI free forever. They don’t have $10B+ free cash to burn and also don’t have that many chips. OpenAI and Anthropic can’t do this either. Only Google, Amazon, Microsoft, Nvidia, Apple and Bytedance have the cash and chips. Unless we have free on-device LLM (which could be possible if Apple behaves), we are talking about tens of billions in inference cost for 100M+ users per year. Then of those with free cash, Only Google, Apple, Amazon, Bytedance have existing distribution that makes CAC free. If Apple can make local LLM great, they might kill meta muse, but only for iphone users. Google, amazon, bytedance aren’t that trustworthy either with user data. Hence, users would just deal with it. People are too lazy and cheap to worry about their privacy. Unfortunate but true.
121
William Chen retweeted
Carnegie report on NeurIPS 2025 authors and AI talent flow. US AI is heavily dependent on attracting talent from PRC. But PRC talent pool is larger than RoW combined. Largest pool of AI talent - capable of publishing at NeurIPS: PRC undergrad → grad school → Chinese company or professoriate. 1. NeurIPS sample undergrad origins: China rose to 57.4% (from 46.3% in 2022); US fell to 13.3% (from 19.8%). 2. Work locations flipped: China now leads at 40.6% (up from 27.1%), US at 34.2% (down from 46.4%). 3. Largest education-to-work path is fully domestic China (undergrad + grad + job): 3,619 researchers, far ahead of China→US→US (1,071) or US→US→US (1,093). 4. US still records the biggest net talent inflow (+2,145); China shows a net outflow (−1,729) even as it produces and keeps far more researchers at home.
Carnegie just dropped its new AI Talent Tracker and the shift since 2022 is pretty striking. Among the NeurIPS researchers in its sample, 57.4% did their undergrad in China, up from 46.3% in 2022. The US fell from 19.8% to 13.3%. But the more interesting change is where they now work
6
18
1
79
7,020
William Chen retweeted
I am done with this shit. It is over. The state of engineering right now is horrible. It has been half a month since I started a new role at a big company. Nobody knows anything here. The specs, code, tests, PRDs, tickets, resolution of those tickets, reports, etc., everything is made by Claude Code. Nobody on my team likes this. They are being forced to ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck, so why are we slow? People are working 12 to 13 hours a day just to press enter. Nobody is reading anything. Humans in corporate are doing nothing on their own. Everyone, literally everyone, from an L1 to an L7 engineer here is doing the same thing. Talk to Claude. There is no sense of victory. Nobody is resolving bugs. In reality, nobody is thinking anymore. Everything is done by LLMs. It is so soul-sucking. I would not mind it, to be honest, if we were at least given the time to check out the code and see what is going where. But no, the goal is to just ship. No matter what happens.
2,049
4,699
1,644
48,616
8,361,117
William Chen retweeted
Fabs are the moat
NEVER thought i would hear this from jeff dean he thinks we can compress chip design from 2 years to 3 months with RL + new EDA tooling essentially just by a specialized auto research loop for hardware!
1
2
1
44
19,900
William Chen retweeted
The funniest thing is he was pro-AI until its capabilities surpassed him
Mathematician Terence Tao: "we have to slow down AI. the pace is insane, and there's no reason to be this fast — no reason at all" It's amazing how willing we are to change everything without any idea what happens afterward These are extremely nonlinear dynamics
383
763
77
14,516
810,898