NEWS YOU CAN’T MISS

Silicon Valley’s Obsession With Pokémon

Want to test how good an AI model is? Make it play Pokémon.

When Pokémon Red and Blue launched in the 1990s, the games became a global obsession. Millions of kids spent hours roaming pixelated worlds, memorizing maps, managing teams, and solving puzzles.

Now, top AI labs are handing AI the controller. Teams at Anthropic, OpenAI, and Google are dropping their models into Pokémon games and watching how they cope. 🎮

Here’s why it works:

• Most AI benchmarks are quick: a question in, an answer out. Pokémon forces hours of planning, memory, and recovery from mistakes.

• It’s based on long-term decisions: where to go, what to train, what to catch, how to manage limited resources.

• The game is full of mazes and puzzles, and progress depends on keeping a goal in mind while navigating lots of small steps.

The approach gained attention when AI company Anthropic launched a live Twitch stream called “Claude Plays Pokémon.” Viewers watched its AI reason through obstacles, explain its choices, and gradually improve over 200,000 in-game steps - offering a look at how AI learns over time.

ps. Some AI models have now beaten the original Pokémon games, and researchers are moving on to tougher sequels. 😲

(Source: Isabelle Bousquette reporting for The Wall Street Journal)

Talking Like an AI Yet?

We already know AI is changing how we write, but this study reveals it is now reshaping how we speak.

After analyzing over 740,000 hours of human speech - including hundreds of thousands of academic talks and podcasts - researchers discovered that humans are unconsciously mimicking the "voice" of AI in real-world verbal communication.

Since the release of ChatGPT, there has been a sudden, measurable spike in humans using specific "AI-favorite" words in conversation. The study highlights "delve" as a prime example, along with words like "comprehend," "boast," "swift," and "meticulous." These words, which LLMs statistically over-generate, are now appearing significantly more often when humans talk to one another.

Meet Barbara Lavernos, Deputy CEO at L’Oréal

Celebrating this week’s Woman in Tech 🥳: Meet Barbara Lavernos, the Deputy CEO helping reinvent L’Oréal - the world’s biggest beauty group, whose products sit on bathroom shelves and dressing tables across the globe.

Barbara Lavernos was born in France and trained as a chemical engineer at Hautes Études d’Ingénieur in Lille. In 1991, she joined L’Oréal on the factory floor, learning firsthand how beauty products are made, sourced, and scaled at an industrial level.

Over the next three decades, she worked her way through some of the toughest, least visible parts of the business: running manufacturing plants, overhauling global procurement, and eventually overseeing L’Oréal’s entire operations and supply chain - including tens of factories and billions of products.

Today, Barbara is Deputy CEO of L’Oréal.

Under her leadership, L’Oréal has pushed hard into beauty tech: AI skin and hair analysis, biotech-based ingredients, smart devices, and science-led personalization. One of the company’s new skin‑health molecules, Melasyl, was selected among TIME’s “Best Inventions 2025.

Lavernos is equally focused on what happens behind the scenes: sustainable packaging, cleaner supply chains, and reducing CO₂ across the entire value chain.

Lavernos believes tech isn’t about chasing the newest tool. It’s about aligning human needs, nature, and industry - and getting everyone, from suppliers to scientists, moving together.

AI Around the World

In the US and Europe, around 800 artists (including Scarlett Johansson, Cate Blanchett, and Breaking Bad creator Vince Gilligan) have signed an open letter accusing AI companies of using copyrighted work to train models without permission, and calling it “theft”.

In Sweden, the National Library of Sweden and the University of Gothenburg are building new AI models trained on Swedish books, newspapers, and speech archives. Backed by government funding, the goal is to ensure future AI tools understand Swedish language, culture, and context more accurately.