Each week, we bring you 5 stories that resonated the most in our internal Slack channel #AI-news. We write the newsletter using various AI tools because we're an AI company and our marketing wants to move with the times too. 😎
Today you're reading the 71st issue in a row.
#1
OpenAI 🚀: Unveils its most powerful AI model yet—o3-pro 🧠📈.
OpenAI 🤖 has introduced its newest and most powerful AI model to date—o3-pro 🚀. It’s a next-generation reasoning model 🧠, designed to solve complex tasks step by step 🔍. That makes it more reliable for math, science, coding 💻, and business use cases 💼. It replaces the older o1-pro in ChatGPT Pro and Team, and is also available via API 🔗.
Experts say o3-pro is better across the board 📈: more accurate, more helpful, and more responsive to instructions ✅. It can browse the web, analyze files 📂, and write code 🧑💻. However, it doesn’t yet support image generation 🖼️ or Canvas, and it’s slightly slower 🐢 than its predecessor.
Despite these limitations, o3-pro outperformed top models from Google and Anthropic in internal tests 🏆—marking a significant step in the race to build the most intelligent AI on the market 🌍🧠.
#2
Meta 💸: Buys stake in Scale AI and hires its CEO to lead AGI efforts 🧠!
Meta 💰 is investing $14.3 billion to acquire a 49% stake in Scale AI and has hired its CEO, Alexandr Wang 🧑💼, to lead a new AI lab focused on developing AGI—artificial general intelligence 🤖. Wang will report directly to Mark Zuckerberg 🧑💻, while Scale AI will continue supplying data to major players like Google, OpenAI, and others 🌐.
The goal is to boost confidence in Meta’s AI capabilities 🔧, especially after the underwhelming launch of Llama 4 🐑. By taking a minority stake, Meta avoids triggering antitrust action ⚖️, while still gaining top-tier talent 🧠 and infrastructure for next-gen AI development.
Meta AI already serves over a billion users through apps like Facebook 📘 and Instagram 📸, but its standalone AI app hasn’t made a major impact yet 📉. This move signals Meta’s clear intent to catch up with AI leaders 🏃♂️—both in cutting-edge development and large-scale adoption.
#3
The Illusion of Intelligence 🤖: Apple exposes the limits of today’s AI 📉.
Apple 🍏 has released a study titled The Illusion of Thinking 🧠, showing that even the best AI models designed for “reasoning” don’t truly understand problems ❌. They perform well on simple tasks but fail when things get more complex—regardless of their computational power ⚙️.
The findings support critiques from experts like Yann LeCun 📚: today’s AI isn’t actual thinking, but advanced pattern recognition 🔁. These models simply predict likely answers based on training data—without real comprehension 🧩.
Why this matters 📌: these systems may appear intelligent, but it’s just an illusion 🎭. Mistaking smooth responses for true understanding can lead us to overestimate what AI—or even people—are actually capable of ⚠️. Apple warns: don’t judge “intelligence” solely by how good an answer sounds 🔊.
#4
Mistral 🧠: Unveils the first European AI reasoning model 🇪🇺.
French 🇫🇷 startup Mistral has introduced Magistral—the first European AI model focused on logical reasoning 🤔. Unlike standard chatbots, it solves problems step by step 🪜, mimicking human thought processes 🧬.
With this move, Mistral joins the ranks of major U.S. 🇺🇸 players like OpenAI and Google, as well as Chinese 🇨🇳 efforts like DeepSeek. At the same time, it stands out by offering open-source models 🆓—a rare trait among top-tier AI systems 💎.
Why it matters 📌: AI development is shifting from simply making models “bigger” to making them smarter 💡. If reasoning models are the future, Mistral may find success even without Silicon Valley–sized budgets 💼.
#5
AI Weather 🤖: Google tests new model to predict storms up to 15 days in advance ⚡.
Google 🌪️ has developed a new AI model for predicting tropical cyclones 🌀 and is testing it with the U.S. National Hurricane Center 🇺🇸. The model can simulate 50 possible storm paths and intensities up to 15 days ahead 📅. In early tests, it was on average 87 miles more accurate 📍 than the leading European forecasting model when tracking storms in the Atlantic and Pacific.
This development is part of Google’s broader effort to make weather forecasting ⛅ an AI-driven process 🤖. Their new Weather Lab platform 🌐 allows side-by-side comparisons of AI forecasts with traditional models—but it’s currently just a research tool 🔬, not an official alert system 🚨.
Why it matters 📌: Climate change 🌍 is making storms more intense and less predictable ⚡, while public weather agencies are facing budget cuts 💸. If AI tools like this prove reliable ✅, they could greatly enhance global warning systems 🌎—without replacing the existing infrastructure.
