Tech Titans Clash: Mira Murati Unveils Thinking Machines Lab, Elon Musk Debuts Grok-3

Mira Murati’s Secret AI Project, Elon’s Grok-3 Challenge, and a Whole Lot of GPU Drama

Alright, buckle up, because today’s AI news is juicier than a watermelon at a summer picnic. We’ve got ex-OpenAI folks launching stealthy startups, Elon Musk going all-in on AI supercomputers, and a high-stakes coding challenge that could change freelancing forever. Let’s get into it!

Mira Murati’s Mysterious AI Startup Finally Steps Into the Spotlight

For months, Mira Murati, OpenAI’s former Chief Technology Officer, has been hiding in the shadows like a ninja, quietly assembling an all-star team of AI experts. Now, she’s finally revealed her new startup: Thinking Machine Labs (which kind of sounds like a science experiment from the future).

Backed by $100 million, Murati’s team—including top brains from OpenAI, Character AI, and DeepMind—wants to build AI that’s more flexible, adaptable, and personalized. What does that mean exactly? Well… we’re not totally sure yet. But with that kind of talent and cash, they might just pull off something groundbreaking—or at least create an AI assistant that can finally understand your grandma’s long, emoji-filled text messages.

Elon’s Grok-3 AI Just Dropped, and It’s a Beast

Speaking of AI power moves, Elon Musk’s xAI just dropped Grok-3, and it’s already stirring up drama in the AI world. The new model is apparently the strongest AI ever, outperforming ChatGPT, Claude 3.5, Gemini 2, and DeepSeek’s latest version. But here’s the kicker—it took one of the largest AI supercomputers ever built (nicknamed "Colossus") to train it!

Elon isn’t stopping there either. His plan? To scale up to 1 million Nvidia GPUs, ensuring Grok-3 (or Grok-4, Grok-5, or whatever comes next) dominates the AI battlefield. But before you freak out about a Skynet takeover—don’t worry, right now Grok-3 is mostly just really good at freelance coding projects and playing Portal 2 (yes, seriously).

Oh, and if you want to access Grok-3? You’ll need X Premium+, which costs a cool $40 a month. So, yeah, another subscription for something else you didn’t know you needed. Thanks, Elon.

OpenAI’s New Coding Challenge: Can AI Replace Freelancers?

OpenAI just rolled out SWE-Lancer, a $1M benchmark test to see how well AI models can handle real-world coding tasks. The test ranges from tiny bug fixes ($50) to full-on feature development ($32,000 projects). Basically, OpenAI wants to see if AI can actually replace freelance coders—or if it’ll just get stuck in an infinite loop and crash your entire project.

Early results? Most AI models are failing HARD. Even the best ones can’t fully take over a coder’s job (yet). So if you’re a developer, you can breathe a little easier—for now.

Other Cool AI News You Can Drop at Dinner Tonight

• **HP just spent $116M buying Humane—**but they’re ditching its infamous AI Pin (probably because no one wanted it).
• A massive AI data center is coming to South Korea, potentially costing $35B when fully built. That’s a whole lot of electricity bills!
• Ilya Sutskever, another ex-OpenAI mastermind, is raising a whopping $1B for his new company, Safe SuperIntelligence Inc. They make AI sound like an edgy sci-fi movie, and honestly, I’m here for it.

2 AI Tools You Probably Didn’t Know You Needed

🚀 Spinach AI – Hate taking meeting notes? This AI automatically records, transcribes, and organizes discussions so you can actually listen instead of pretending to while furiously typing.

📊 Basejump – Ask a data question and get answers in seconds instead of digging through spreadsheets for hours. Perfect for when your boss asks those “quick data insights” at the most inconvenient time.

Alright, that’s all for today’s AI adventure. If you want more wild updates like these, follow us on Twitter here. 🚀