What if one of Tesla’s most overlooked AI infrastructure inventions is not a chip, but a file format? 👀 It is called ".smol" 💥 At supercomputer scale, the bottleneck is not always the GPU. These massive math engines are fast, but they can still end up waiting while the system parses, decodes, scans, and reshapes training data ⏳ Tesla’s patent WO 2024/073080 attacks that bottleneck at the root: the file itself 🗃️ Instead of treating training data like a sequential file, .smol turns it into a random-access machine. Incredibly, this architectural shift can cut I/O operations by roughly 4x and trims file size by about 11% 🆒 If you don’t have time for the full deep dive, here is how the invention works in 60 seconds: 🚀 The header becomes a map Every timestamp or index is mapped to a precise byte offset inside the file. Instead of crawling through the data like an ancient scroll, the reader checks the header and jumps straight to the exact row it needs. 📦 Static data stops being copied everywhere Details that do not change, such as vehicle dimensions, calibration settings, or other constant metadata, can be stored once before the timestamped rows. The file stops wasting space repeating the same background information again and again. 📊 The row is arranged for cheap decisions first Lightweight fields like velocity or steering angle can sit near the front, while heavier tensors or video-derived chunks sit later. That lets the system read the “headline” first and avoid touching the expensive payload when it is not needed. ⏱️ Each timestamp becomes a sensor bundle Multi-camera and sensor data can be grouped around the same timestamp or index, so the training system retrieves aligned views of the world instead of stitching together mismatched streams. 🧠 The file speaks the model’s language The format supports tensors, the mathematical structures neural networks actually consume. Some tensors can also be stored already transposed into the shape the model expects, reducing the work needed before the data reaches the GPU. 🔒 The layout stays fixed Because the file can be read-only, the byte offsets do not shift underneath the index. That makes the header reliable: the address still points to the same row tomorrow that it pointed to today. 🤖 The same idea scales beyond cars The patent defines “ego” broadly, covering vehicles, trucks, and even bipedal humanoid robots. So .smol is not just about FSD video clips; it points to a general data infrastructure for machines that perceive and move through the world. The big idea is simple: Tesla is not only trying to make the GPU faster. It is trying to make the data arrive in the shape, order, and location the AI system needs. At AI scale, the bottleneck is not always the brain. Sometimes it is the pipe feeding the brain.

There are currently 16 public companies worth more than $1 Trillion 🥇 Nvidia $NVDA - $5T 🥈 Google $GOOGL - $4.4T 🥉 Apple $AAPL - $4.3T 4: Microsoft $MSFT - $2.9T 5: Amazon $AMZN - $2.6T 6: Taiwan Semi $TSM - $2.2T 7: SpaceX $SPCX $2.1T 8: Broadcom $AVGO $1.8T 9: Saudi Aramco - 1.8T 10: Tesla $TSLA - $1.5T 11: Meta Platforms $META - $1.4T 12: Samsung - $1.4T 13:Micron $MU - $1.1T 14: Berkshire Hathaway $BRK.B $1.1T 15: Eli Lilly $LLY - $1T 16: SK Hynix - $1T

Ten years ago, Elon Musk and Jack Ma both looked like avatars of the future. Musk was building electric cars, rockets, satellites, and AI. Jack Ma was building Alibaba, the crown jewel of China’s internet economy. Fast forward a decade: Musk is worth $1.1 trillion. Jack Ma was forced into silence, Ant Group’s IPO was crushed, Alibaba lost its aura(its stock plunged by more than 60% since 2020), and China’s tech sector went from world-beating to politically domesticated. One entrepreneur was allowed to compound. The other was reminded who really owns the scoreboard. This is not just a story about two billionaires. It is a story about two systems. One system lets madmen build rockets to Mars. The other cuts down its most successful entrepreneur for flying too close to the sun. That is why Musk became the world’s first trillionaire. And Jack Ma became a warning label.

Congratulations @MrBeast on making history as the first creator to reach 500 million subscribers on @YouTube! 🎉 This achievement is a testament to what’s possible when a creator has the freedom to follow their creativity and be their authentic selves. Jimmy, you are an incredible innovator and we're so proud to be part of your journey.

Protesters are standing outside JPMorgan Chase in New York this weekend, holding giant banners that say “STOP ELON • NO TRILLIONAIRES” and “ELON IS STEALING YOUR PENSION” These people are protesting the greatest success story in modern history. Let’s look at the actual reality: • Elon has already paid more in taxes than almost anyone alive • Tesla and SpaceX employ well over 100,000 people and actively push humanity forward. • Through DOGE, he already helped cut billions in wasteful government spending. That saved taxpayers hundreds of billions.....meaning he actually helped secure grandmas' pensions by slashing bureaucracy, not the other way around. • Friday's historic SpaceX IPO didn't "steal" from anyone. It created massive value, turned everyday workers into millionaires, and rightfully made him the world’s first trillionaire He didn’t steal anything. He built it from the ground up His companies have created massive job growth, accelerated sustainable energy, and are literally building the infrastructure to make humanity multi-planetary Rather than celebrating excellence, demanding the government punish it is pure evil

Buffett seeks to win by finding moats, which usually arise out of competitive mkt failures. Great for Buffett, who gets to extract higher rents, but not for the rest of society, which has to pay them. Musk seeks to win by innovating faster & better than his competition. Great for everyone. In his efforts, he has created several orders of magnitude more value for the world than Buffett. There really is no comparison.

People are confused about Jeff Bezos' new startup Prometheus that just came out of stealth with $12 billion in funding and a $41 billion valuation. When news broke about the venture last year, many assumed it was the Amazon and Blue Origin founder's foray into the physical AI goldrush. Others thought it was yet another LLM. A few weeks ago, Bezos said it was more like a "very advanced" version of CAD software. But there are a million Text to CAD startups right now, and none of them have funding anywhere near $1 billion, let alone $12 billion. Bezos and Prometheus co-founder Vik Bajaj, a former Google/Veriliy exec, gave the most detailed explanation of what they're doing this in an interview with CNBC. Bezos said they're developing an "artificial general engineer" to design, test, simulate, and manufacture complex physical products like jet engines, spacecraft, medical devices, consumer electronics, industrial machinery, and buildings. He said the goal is to shorten the "invention loop," accelerating the path from idea to finished product. "The idea that you might build a set of tools that can actually do engineering, an artificial general engineer, it's a dream that we've had ... for decades," he said. "But it's never really been possible." The technology, as described, could be disruptive several layers of industry at once: - CAD/PLM and digital engineering software companies like Autodesk, Dassault Systèmes, Siemens, Ansys, Altair, and Hexagon AB; - Industrial automation and manufacturing software players like Siemens, Rockwell Automation, Honeywell, Schneider Electric, and Emerson - Manufacturing execution systems (MES) like Siemens Opcenter, Rockwell FactoryTalk, ProductionCentre, DELMIA, SAP Digital Manufacturing, Tulip, and AVEVA The most disruptive version of Prometheus would sit across design, simulation, sourcing, production planning, testing, quality control, and factory optimization. The way Bezos talks about it makes it seem like it could compress years of engineering iteration into months or weeks. He told CNBC that Prometheus and other next-gen AI tools will lead to a labor shortage, not mass unemployment. "Six thousand years ago, somebody invented the plough and we all got wealthier. Much later, somebody invented the steam engine and we all got wealthier," he said. "These things drive productivity."

SpaceX raised only $12B of capital before going public. With that $12B, they revolutionized the rocket industry, built a global satellite network, and created arguably the most innovative company of all time. The federal government spends $12B every 15 hours and still can’t get its shit together. Prior to SpaceX, NASA was sending astronauts into space on Soviet-era Russian Soyuz capsules. So no, I don’t find Elon’s wealth to be a problem, and I wouldn’t trust Elizabeth Warren or Bernie Sanders to allocate a single dollar of it.

I’m just getting home after two days at Starbase It is impossible to put into words the work ethic and mission orientation of the team at @SpaceX Even on IPO day they launched a rocket, were doing engineering stand ups and welding new super structures They’re doing this for the betterment of humanity There’s nothing else like it

About 5 years ago, Elon was giving @Erdayastronaut a tour of the Starbase launchpad. A worker at the site walks up, hugs him, and says: “We’re going to make it.” He’s probably a millionaire now.

Gwynne Shotwell joined SpaceX in 2002 as its 11th employee. Today, as president and COO, she oversees operations at the commercial space exploration company. Shotwell is featured on Forbes Richest #SelfMadeWomen. Read the full list: forbes.com/lists/self-made-w… Photo: Angel Garcia via Bloomberg

I found the SpaceX IPO surprisingly emotional today. It reminded me of why I came to the US as an 18-year-old. Even then, I knew that I wanted to build stuff, and that the place to do it was America. There was no second option in the world. The concept of American exceptionalism is nothing new, but I have come to appreciate the culture that justifies it. SpaceX is yet another case in point. America’s risk-taking culture celebrates wild successes while embracing the legitimacy of hard-earned failure. American culture doesn’t celebrate inherited wealth, nor does it frown upon inherited poverty. It doesn’t seek to create equal outcomes, but rather equal opportunity. It is not immediately obvious that this is unique in the world. Truly unique. I’m Canadian, and I love many aspects of Canadian culture, but Canadian culture does not offer people the same environment in which to take risks. I write this because I was dismayed to see US politicians complaining of the extreme wealth created by the SpaceX IPO. Say what you will about wealth inequality, or a single man’s politics, but don’t tell me that immense wealth creation in America is bad. Do not tell me you’d rather SpaceX not exist, exactly as it does. That you’d rather this company exist in some other country or culture. Thankfully, despite today’s politics, SpaceX could not have been a Chinese company, or a Canadian one, or a French one. It could only ever have been an American one. If nationalism is pride in your birthplace, then it’s merely tribalism, which serves to divide us. But if it’s pride in your culture, a culture that lets people achieve incredible things like this, then under those terms I am a nationalist. I want to protect and enhance our culture of risk-taking, of celebrating wins, and of celebrating failures along the way. I think it is amazing that America created a trillionaire out of a risk-taking immigrant. It is absolutely fucking absurd, of course, but isn’t that the point? SpaceX is not a reason to be pissed off; it’s a reason for every person in the world who wants to build stuff to see themselves as American, no matter where in the world they live. p.s. — this is entirely from my brain, with AI used only for fixing typos and grammar. :^)

Looks like the $SPCE buyers realized their mistake. Yeah I don't believe in the efficient market hypothesis...