We were never meant to fade quietly. Build the machines. Light the engines. Reach for the impossible. Rage, rage against the dying of the light. "Final Surge" Music Of My Creation

Elon Musk’s real masterclass isn’t rockets, cars, or AI. It’s problem-solving. > Strip away assumptions. > Ignore “that’s how it’s always been done.” > Reduce the problem to what is physically true. > Then rebuild the solution from first principles. Common sense + first principles is a ruthless combination. Most problems look complicated because we inherit the complexity. Start from zero.

Humanity is ~Type 0.7. > Type I: master planetary-scale energy + orbital compute. > Type II: industrialize the Moon and surround the Sun with a Dyson swarm. > Type III: replicate that architecture across a galaxy. The jump from us to K3 is ~20 orders of magnitude in power. The ladder is simple: Reusable launch → orbital energy → off-world industry → self-replicating AI → the stars. K3 isn’t one invention. It’s infrastructure that learns how to reproduce itself.

This may be one of the most important Starship recoveries yet. After Flight 13, Ship 40 completed its reentry and soft splashdown in the Indian Ocean then did something previous Starships hadn’t: it stayed largely intact. SpaceX eventually recovered the vehicle, giving engineers direct access to hardware that had just survived ascent, hypersonic reentry and landing. And they found things telemetry alone couldn’t fully show. Post-flight inspection identified areas where plasma could find flow paths behind heat-shield tiles. SpaceX has since added retention mechanisms in higher-risk areas and is testing curved tiles designed to reduce heating in the gaps. Even more important: two tiles recovered from Ship 40 are planned to fly again on Ship 41. That is the real prize. Starship’s hardest problem isn’t simply surviving reentry once. It is building a thermal protection system that can survive reentry, be inspected minimally, refueled and flown again. A soft landing gave SpaceX the hardware to dissect instead of just the telemetry. And somewhere in those scorched tiles, attachment points and plasma pathways may be the key to true rapid reuse.

Simply Explained Ep. 01: How Does SpaceX Fuel Starship? You see all those giant pipes and equipment connected to Starship? Think of them as part of the rocket’s giant gas station. Starship needs two main liquids: liquid methane, which is the fuel, and liquid oxygen, which helps the fuel burn. Both are kept extremely cold. Before launch, SpaceX pumps huge amounts of these liquids from massive storage tanks on the ground, through large pipes, and into Starship. That big arm connected to Starship is called the Quick Disconnect Arm. Think of it like a giant fueling and charging cable. It supplies Starship with fuel, oxygen, electricity, and data while the rocket is on the launch pad. Just before launch, the arm disconnects and quickly moves away from Starship. Then the engines ignite, and Starship lifts off. Simply explained: all those giant tanks, pipes, and equipment connected to Starship work together like a massive gas station built for a rocket.

Elon Musk’s TeraFab vision gets wild fast: AI in space could become cheaper than AI on Earth within 2–3 years. Why? Near-constant sunlight, 5×+ solar yield, less battery mass, fewer terrestrial power constraints then scale toward a terawatt of orbital compute. And eventually? Lunar mass drivers. Petawatt-scale compute. Millions of tons to orbit. This isn’t just about building better chips. It’s about moving civilization’s compute infrastructure off-world.

The most interesting thing about Terafab isn’t how many chips it could produce. It’s what happens when the feedback loop collapses. Today, building an AI processor means coordinating across an enormous distributed supply chain: architecture → foundry → memory → packaging → testing → systems → real-world deployment → new architecture. Every handoff adds latency. Terafab attacks the latency itself. Tesla and SpaceX are building toward vertically integrated logic, memory and advanced packaging because the amount of silicon required for autonomous vehicles, humanoid robots and eventually space-based compute may exceed what the existing semiconductor ecosystem was designed to supply. That creates something much more important than manufacturing capacity: a hardware learning loop. Deploy AI5. Collect real-world inference data. Find the bottlenecks. Change the architecture. Change the memory hierarchy. Change the packaging. Build the next generation. Repeat. Tesla says AI5 has already completed final chip design, with AI5 and AI6 production planned for 2027 and 2028. If Terafab reaches enormous wafer scale, the strategic advantage won’t simply be owning a chip factory. It will be shrinking the distance between learning something about intelligence and physically encoding that lesson into the next generation of silicon. Software already iterates this way. Terafab is an attempt to make hardware iterate more like software. That is the part worth watching.

SpaceXAI is testing Remote Control for Grok Build This is one of the biggest Grok Build updates yet....and the one future I was desperately waiting for Grok Build... it's finally coming You’ll be able to link your computer once through the Grok Build CLI, then control Grok Build remotely from the web and directly from the Grok app on your phone That means Grok Build can keep working on your actual computer while you’re somewhere else with nothing but your phone

The New York Times, Dec. 8, 1903: “Man won’t fly for a million years, maybe even 10 million years.” SpaceX: “Flying? That’s done. Multiplanetary is next. Multistellar comes after that.”

Hollywood would spend 7 figures on this Odyssey scene. @PJaccetturo and his studio made this pilot in 9 days, with 3,627 images, 3,043 videos, and generations costing $2,677. Here’s their breakdown of how they did it 🧵