Andrej Karpathy casually shows how to build real apps just by prompting in under 30 minutes. Save this, I promise you'll come back to it.

Forget the $200,000 degree. In 52 minutes, Steve Jobs teaches more about vision, product and the future than a 4-year MBA. This Interview from 1990 predicted the next 35 years of technology with terrifying accuracy. Save this rare footage.

Elon Musk's xAI has just achieved something unprecedented in the AI industry. Colossus 2, xAI's supercomputer, is now operational with one full gigawatt of power capacity, making it the first gigawatt-scale AI training cluster in the world. To visualize this, one gigawatt powers roughly 750,000 homes and exceeds San Francisco's peak electricity demand. The facility will upgrade to 1.5 gigawatts in April and eventually reach 2 gigawatts total, consuming as much power as most major US cities. Inside sits approximately 555,000 specialized computer chips called GPUs that work together to train AI models.​​ This matters because in AI, bigger infrastructure equals smarter AI. An AI model's intelligence and capability scale directly with the computing power used to train it. More GPUs and power allow companies to train larger, more capable models faster. The real advantage is speed, Colossus 1 went from dirt to fully operational in just 122 days. While competitors like OpenAI and Google are still planning their 2026-2027 infrastructure, xAI is already operating at city level power consumption today. Whoever trains AI models fastest can iterate faster and deploy improved versions before competitors catch up.​ The scale is overwhelming. xAI's 555,000 GPUs far exceed Meta's estimated 150,000, Microsoft's roughly 100,000, and Google's distributed infrastructure. The planned 2 GW facility would be approximately 4 times more powerful than the next largest dedicated AI training site. This massive compute capacity will train Grok, xAI's AI model integrated with X's hundreds of millions of users providing both raw computing power and built-in distribution.​ xAI recently raised $20 billion specifically to accelerate this infrastructure and train new versions of Grok. The company ended 2025 with over one million GPU equivalents across Colossus facilities.​​ No one else can do this other than @elonmusk and the team at @xai. Well done!