🥇 AI wins Gold at World's Toughest Math Olympiad

Welcome, AI enthusiasts

AI just went head-to-head with the world’s top mathletes. Both OpenAI and Google hit gold-level scores at the 2025 International Math Olympiad. But here’s the kicker: their AI models didn’t need any help from humans. For the first time, they read the problems and wrote the full solutions all by themselves. Let’s dive in! 

In today’s insights:

• AI wins Gold at World's Toughest Math Olympiad

• Stargate Texas data center comes online

• Meta Rejects EU AI Pact

Read time: 3 minutes

LATEST DEVELOPMENTS

OPENAI & GOOGLE
🎓 AI Takes On the Math Olympiad

Evolving AI: OpenAI and Google both claim gold-level scores at the 2025 International Math Olympiad.

Key Points:

• OpenAI and Google DeepMind each scored five out of six on the 2025 IMO using AI models that read and answered in natural language.

• Both models outperformed most human competitors and improved on Google's 2024 silver medal model, which needed manual translation.

• Tension flared as Google criticized OpenAI for announcing its results before the official IMO evaluation.

Details:

OpenAI and Google have both pushed their AI systems into new territory: solving high school-level math problems at International Math Olympiad standards. Unlike last year, where AI models required human translation, both teams submitted "informal" models capable of reading problems and writing full solutions on their own. The AI models each answered five of six questions correctly, a score worthy of a gold medal. Still, like with most announcements between OpenAI and Google these days, it sparked friction. Google accused OpenAI of jumping the gun by going public before the IMO had officially reviewed its submission.

Why It Matters:

These models didn’t just solve Olympiad problems. They did it in plain language, without custom tools or translations. That is big. It means the AI understood the question, planned a solution, and wrote out a full proof like a person would. This came from smarter architectures, heavy training on proof strategies, and more thinking time per question. It’s a step toward general-purpose reasoning that could help in fields where deep logic and clear explanation are key. Now more than ever, it's time to watch this closely, because with AI reaching this level of thinking and reasoning, we may be on the verge of new discoveries.

OPENAI
🚀 Stargate Texas data center comes online

Evolving AI: OpenAI says its Stargate I facility in Texas is already partially operational.

Key Points:

• OpenAI, Oracle, and SoftBank are expanding their Stargate AI data center project to 5+ gigawatts of capacity.

• Parts of the first data center in Abilene, Texas, are already running, with Nvidia GB200 racks installed.

• The full project aims to add 100,000 jobs and reach 10GW total compute capacity.

Details:

OpenAI and Oracle announced a major expansion of their joint AI infrastructure effort with SoftBank. The Stargate initiative now targets 5+ gigawatts of data center capacity, with construction underway in Abilene, Texas. OpenAI confirmed that parts of Stargate I are already live, powered by Nvidia's GB200 racks. The broader plan is to build 10GW of AI-ready capacity across the U.S., employing tens of thousands.

Why It Matters:

With parts of Stargate already running, we’re about to get faster, smarter, and more responsive AI. That shift will reach everything from coding tools to agents in your browser. The only thing holding it back now is whether the supply of compute can keep up with the demand.

META
🗳️ Meta Rejects EU AI Pact

Evolving AI: Meta refuses to sign the EU’s voluntary AI code, calling it a step in the wrong direction.

Key Points:

• Meta says the EU’s AI guidelines go beyond the law and create legal uncertainty.

• The company fears the rules will slow AI development and hurt European innovation.

• OpenAI has agreed to sign the same code that Meta is rejecting.

Details:

Meta has decided not to join the EU’s voluntary code of practice for general-purpose AI models, arguing the guidelines add confusion and overreach beyond the AI Act. The EU sees the code as a way to smooth the transition before new AI laws take effect on August 2nd. Meta believes the current direction risks holding back progress in Europe. OpenAI, by contrast, has signed on.

Why It Matters:

A clash between American tech and EU regulation is nothing new. Meta stepping away from the EU’s voluntary AI code shows how far apart EU regulators and major American AI companies still are. The code was meant to guide responsible development ahead of the AI Act, but Meta sees it as vague and a threat to innovation. If others follow, Europe may struggle to attract or keep advanced AI systems. That could leave EU businesses and researchers behind, stuck with fewer tools. But is slower future access to AI a fair trade-off for stronger protections for EU citizens?

 👀 Click on the image you think is real

QUICK HITS

🤝 UK and ChatGPT maker OpenAI sign new strategic partnership.

🥰 Most teens have used AI to flirt and chat.

✨ X Plans To Launch AI Text-To-Video Option.

📊 Chinese startup Manus challenges ChatGPT in data visualization.

💰 Anthropic CEO Says the Company Will Pursue Gulf State Investments After All.

📈 Trending AI Tools

• 🌙 Dream Recorder – DIY gadget that turns spoken dream descriptions into low-fi videos for introspective reflection (link)

• 🧬 AlphaEvolve – DeepMind’s Gemini-powered agent that designs and refines algorithms autonomously (link)

• 🏗️ Base44 – Natural‑language, no‑code platform that creates full-stack web/mobile apps conversationally (link)

• 🧮 Lumio AI – Multi‑LLM workspace for comparing ChatGPT, Gemini, Claude, Grok, etc., and building custom agents (link)