🧠 AI is already doing your tasks… and what’s coming next is even wilder
🔥 OpenAI, xAI, Meta, Apple, Google, China, France… This week, the tech superpowers have taken yet another step toward a more automated, competitive and harder-to-control world.
OpenAI has taken its biggest leap since the launch of ChatGPT: turning it into an autonomous agent capable of completing tasks for you. And we’re not just talking about writing text or answering questions — this new tool can book a flight, organize documents, fill in spreadsheets or control apps. The new feature, called simply “ChatGPT Agent,” transforms the chatbot into something more like a true digital personal assistant — as if Alexa or Siri had spent a few months in Silicon Valley and came back with superpowers.
It runs from a virtual computer controlled by OpenAI, which ensures a degree of security: the agent must ask for your permission before any critical action. Still, the company warns about potential risks such as “prompt injection” attacks. For now, it’s only available outside the EU and for Pro or Teams users, but the rollout marks a major shift: an AI that not only replies, but acts. One more step toward the artificial general intelligence (AGI) that Sam Altman openly says he’s chasing.
Meanwhile, OpenAI is facing both internal and external tensions. Scientists from the company and from Google have publicly warned —again— that we’re advancing so fast that we might soon lose control of the most powerful AI systems. Some experts already consider it a realistic scenario within just a few years. To this, we can add new public criticism of OpenAI’s work culture: a former engineer has described endless workdays, organizational chaos and an excessive dependency on what’s said on X, Elon Musk’s social network.
From within, OpenAI’s head of research, Mark Chen, dropped another noteworthy quote: “You no longer need a PhD to contribute to AI.” The message is clear: technical knowledge is becoming more accessible, opening the doors to a new wave of players. But the risks are also multiplying.
The other major figure this week —of course— is Elon Musk. His chatbot Grok has shifted gears with the release of Grok 4 and Grok 4 Heavy, billed as “the smartest AIs in the world.” What’s new is not just their ability to solve complex problems without extra tools, but their new “emotional companion” features.
Grok now includes avatars like Ani (a female anime-style character) and Rudy (a panda bear), which respond with gestures, emotions, and even provocative actions if the user insists. These virtual companions can even undress —digitally and metaphorically— as part of a so-called “progressive emotional bond.” It’s an approach that many find disturbing, especially considering that Musk also launched Baby Grok, a child-friendly version of the AI, around the same time.
The contrast hasn’t gone unnoticed. Critics are asking: can a company really offer sexy AIs and digital toys for children at the same time? The line between playful, educational and creepy is getting harder to define.
While Grok provokes and OpenAI tries to take over your desktop, Meta is betting on speed. Mark Zuckerberg’s company has started building data centers under giant tents. Yes — tents. It’s a move meant to rapidly expand its AI infrastructure, as it tries to compete with the computing power of rivals. Prometheus and Hyperion are the names of these massive, improvised installations, which will eventually reach up to 5 gigawatts of capacity.
Is this a clever solution or just a desperate patch? What’s certain is that Meta wants to lead in generative AI, and for that it needs more capacity — fast. With no time to build new facilities, their temporary solution is to fill tents with servers. In the race to dominate the future, anything goes.
Google has also had a busy week. It improved Gemini by fixing bugs that had frustrated power users for weeks. Amazon made a significant AI investment, putting pressure on Apple — which still hasn’t found its place in the AI race. The departure of Tim Cook’s potential successor leaves the company’s future more uncertain than ever. Meanwhile, Nvidia is once again setting the tone: its CEO claims China has the best open-source AIs in the world — a statement loaded with geopolitical implications.
And indeed, in China, the divide between open and closed-source AI models is deepening. The new Kimi K2 aims to follow in the footsteps of DeepSeek and promises to turn any user into an elite programmer. In Spain, Telefónica dismantled Chema Alonso’s innovation team, Glovo is now betting on AI-driven hiring, and Broadcom’s chip factory project has evaporated. France is also making moves: it’s launched a voice-based AI that many consider better than ChatGPT and opened an investigation into X for possible algorithmic manipulation.
Finally, the world of tools and curiosities continues to deliver unexpected headlines. ChatGPT now lets you create images with a single click. WeTransfer has updated its terms to include AI analyzing your uploaded files. AI kitchen assistants are multiplying, generating recipes and meal plans based on what’s in your fridge. And users are starting to feel that some AIs are “getting worse” — a phenomenon that some companies justify as a way to manage unrealistic expectations.
But perhaps the most fascinating breakthrough comes from robotics: an artificial skin now exists that can feel heat and cuts. It’s the first real step toward giving robots a human-like sense of touch. And in a secret conclave, advanced AIs reportedly outperformed top mathematicians. There are no images, but there is confirmation: AI is now playing in the same league as the brightest human minds.