🧠 OpenAI’s Math Monster Bags Olympiad Gold
Okay, this one’s wild. OpenAI’s latest model just went head-to-head with the actual International Math Olympiad... and snagged gold. It cracked 5 out of 6 problems - real, brutal problems - under official student conditions. No side notes, no tweaks. Straight up.
Sam Altman called it “a milestone,” which yeah, fair. People are already whispering AGI, but let’s not do that yet. Still, math this deep? That’s new. This isn’t just pattern-matching anymore; it’s structured thought. Some critics want third-party verification. Others just seem nervous.
💸 Thinking Machines Gets $2 Billion, No Product - Yet
You know those startups that drop outta nowhere, raise billions, and say, “just wait”? That’s Thinking Machines Lab right now. Led by Mira Murati (yep, ex-OpenAI), they just scored $2B in funding. Seed round. Valuation? $12B.
They’re promising something open-source, agentic, deeply multimodal, and “built for researchers.” Whatever that means. So far? No product. Just buzz and a waitlist. But investors clearly think there’s heat behind the curtain.
🇬🇧 UK Finally Flexes Its Supercomputer Budget
Britain’s finally decided to play the AI infrastructure game for real. They just announced a £1 billion plan to stitch together their existing supercomputers (Isambard-AI, Dawn, etc.) into one mega-grid - the “AI Research Resource.” It’ll boost compute twentyfold. That’s not a typo.
They’re partnering with Nvidia, Intel, Dell… basically all the hardware players. Part of it’s national pride. Part of it’s real science: climate modeling, drug design, fusion. Feels like they’re tired of playing second string to the U.S. and China.
🔌 Nvidia’s China Reboot? Delayed, Again
Nvidia’s plan to restart its H20 chip sales into China just hit a wall, again. The production line isn’t ready. TSMC had to reconfigure after U.S. export restrictions, and now we’re looking at another 6–9 months of delay.
So what’s Nvidia doing? They’re pivoting to the RTX Pro, a slightly dialed-down version cleared by regulators. It’s not ideal. But it’s all they’ve got. Another reminder: AI hardware isn’t just specs and silicon, it’s borders, approvals, and geopolitical chess.
📈 U.S. Startups Rake in Cash, But Funds Are Drying Up
U.S. AI startups just posted a monster first half: $162.8B raised, up 76% year-over-year. That’s the biggest H1 since 2021. Everyone from Safe Superintelligence to dozens of stealth-mode agent labs got a slice.
But here’s the catch, VC fundraising is down 33%. Which means? There’s still money flowing into startups, but fewer new funds forming to spread it around. The capital’s consolidating. It’s getting quieter out there, except at the top.