!
👋 Dear Dancing Queens and Super Troupers,
This CES week, AI didn’t just put on a show. It cracked open the living-room door, stepped into the kitchen, peeked into our grocery carts… and started poking around inside our brains.
On one side, the promises of comfort are almost indecent. OpenAI tells us its future agents will help manage household chores. Making lists, organizing, optimizing, delegating. Less mental dishwashing, more free time.
Google, meanwhile, is rolling out the red carpet for agentic commerce with a protocol designed to let AI buy things for us directly. No clicks, no websites. Carrefour is already on board. The shopping cart becomes conversational. The consumer, optional.
At the same time, in CES labs and closed-door demos, robotics is accelerating. Agibot unveils Genie Sim 3.0, a massive machine for training artificial brains inside simulated worlds. Robots that learn before they even physically exist.
Humanoids repeating the same gestures thousands of times, while we humans repeat the same routines. AI advances through massive simulations. We advance through habits.
Then there’s the other side of the story. The one that scratches beneath the surface. The MIT publishes an unsettling study: heavy use of generative AI could reduce cognitive engagement, mental effort, and our ability to remember what we produce.
Not because AI is “bad.”
But because it’s too effective when we let it think for us.
At the same time, Google urgently pulls certain health-related AI summaries after dangerous medical responses. Here, we’re no longer talking about comfort or productivity. We’re talking about bodies, real risks, misplaced trust.
AI no longer just assists our screens. It’s starting to structure our decisions, our actions, our everyday choices. It saves time, yes. It smooths friction, without a doubt. But it also subtly shifts effort, responsibility, and attention.
So the real question is no longer “Is AI impressive?”
We’ve moved past that.
The question now is: what do we still choose to do ourselves, and what are we willing to delegate without thinking? Because if we automate everything, we might end up stuck in permanent energy-saving mode… brain included.
Here’s this week’s lineup :
👉 Thinking less thanks to AI: does comfort come with a cognitive cost? 🧠
👉 Medicine and AI: one mistake too many for Google ⚕️
👉 Agibot bets on simulation to accelerate embodied AI 🤖
👉 Cleaning, organizing, chores: OpenAI targets everyday life 🏠
👉 Google wants AI to do your grocery shopping with Carrefour 🛒
If we've forwarded this letter to you, subscribe by clicking on this link !
⚡ If you have 1 minute
An MIT study shows that passive use of tools like ChatGPT reduces cognitive engagement, memory retention, and personal satisfaction. AI itself isn’t the problem, but when it replaces effort instead of supporting it. The risk isn’t the technology, but the gentle dependency it creates.
After incorrect and potentially dangerous medical recommendations, Google removed its “AI Overviews” for certain sensitive health queries. The episode is a blunt reminder that generative AI is not neutral when it comes to bodies and diagnoses. Here, errors are no longer abstract.
Presented at CES 2026 by Agibot, Genie Sim 3.0 is an open-source, large-scale simulation platform designed to train robotic AI before it physically exists. Thousands of hours of data, hundreds of tasks, one clear goal: accelerate embodied AI at full speed. The future of robots is being shaped first in virtual worlds.
According to OpenAI, AI agents will soon help organize, plan, and manage household tasks. A form of invisible productivity, absent from GDP metrics but very real in daily life. AI no longer promises to “change the world,” it mostly promises to free us from the dirty work.
Google is launching an open protocol that allows AI assistants to purchase directly from merchants, without traditional interfaces. Carrefour is among the first partners. The shopping cart becomes conversational, the act of buying programmable. The human buyer starts sharing the steering wheel.
🔥 If you have 15 minutes
1️⃣ Thinking less thanks to AI: does comfort come with a cognitive cost ?
The summary : Using generative AI to work faster is not without consequences. A study from the Massachusetts Institute of Technology (MIT) shows that passive use of ChatGPT reduces mental engagement, weakens memory, and slows the development of cognitive skills. A short-term comfort gain that could come at a high cost to the brain in the long run.
The research team measured brain activity in real time, rather than relying on simple questionnaires, by asking participants to write SAT-style essays with or without AI assistance.
Details :
A less engaged brain : ChatGPT users showed significantly lower neural activity than those working without assistance, indicating reduced cognitive effort.
The habituation effect : The more AI was used, the more mental engagement eroded. Some participants admitted to copying generated text with little personal involvement.
Convenience has a price : Compared with traditional search engine use, AI drastically simplifies the task but creates a measurable form of cognitive dependency.
Work that loses its value : AI-assisted participants reported lower satisfaction with their output and struggled to quote their own writing, due to a lack of deep processing.
“Cognitive debt,” according to ChatGPT : The AI itself acknowledges that excessive use as a mental crutch can lead to a gradual atrophy of reasoning abilities.
Why it's important : Published on arXiv under the title Your Brain on ChatGPT, the MIT paper confirms a warning already raised earlier this year by Microsoft about the erosion of critical thinking. AI remains a powerful accelerator, provided we don’t let it think in our place.
2️⃣ Medicine and AI: one mistake too many for Google
The summary : Google has quietly removed its AI-generated overviews for certain sensitive medical queries following an investigation published on January 11, 2026, by The Guardian. The issue: responses deemed misleading about health test results, potentially putting users at risk. A swift decision, but one that raises broader concerns about the use of AI in medical search.
Details :
An investigation that triggered the alert: The Guardian revealed that Google’s AI overviews provided “normal” ranges for liver tests without accounting for key factors such as age, sex, nationality, or ethnic background.
Targeted medical queries: AI summaries disappeared for searches like “normal range for liver blood tests” and “liver function tests.”
Variants still active: Despite the removal, closely worded queries such as “reference range for liver tests” initially continued to display AI summaries.
A discreet but real rollback: Hours after the article’s publication, multiple tests showed the absence of AI overviews, even though Google still offered to run the query in AI mode. In some cases, The Guardian article itself appeared as the top result.
Google’s official response: A spokesperson declined to comment on individual removals, instead citing “broader improvements.” An internal team of clinicians reportedly judged many of the responses to be accurate and supported by reliable sources.
Why it's important : As early as 2025, Google had promised AI models dedicated to healthcare. This episode shows that in medicine, an algorithmic approximation isn’t just a minor bug—it represents a real risk for users.
3️⃣ Agibot bets on simulation to accelerate embodied AI
The summary : Unveiled at CES 2026, Genie Sim 3.0 marks a new milestone for embodied intelligence. AGIBOT introduces an open simulation platform, integrated with NVIDIA Isaac Sim and NVIDIA Omniverse. It is designed to train, test, and benchmark intelligent robots faster, with less hardware, and with large-scale, finally standardized benchmarks.
Details :
An ocean of open data : A unified open-source platform for advanced robotics, AGIBOT integrates more than 10,000 hours of realistic synthetic data, including RGB-D sensors, stereo vision, and full kinematic data.
An all-in-one pipeline : The platform brings together digital asset creation, scene generation, data collection, physical simulation, and automated evaluation within a single environment connected to NVIDIA Isaac Sim.
Fewer robots, more iterations : According to AGIBOT, Genie Sim 3.0 significantly reduces reliance on physical robots while accelerating model training cycles.
Large-scale benchmarks : The Genie Sim Benchmark evaluates embodied intelligence across more than 200 tasks spread over 100,000 simulated scenarios, producing comprehensive capability profiles.
Language-driven scene generation : Using large language models, users describe an environment in text. The system generates scenes, semantic variations, and visual previews without manually coded logic.
Industrial digital twins without hardware : Genie Sim 3.0 reconstructs digital twins of logistics sites and production lines, enabling full testing without deploying a single robot
Why it's important : By making data, simulations, and evaluation tools fully open source, AGIBOT aims to lay the groundwork for future standards in embodied AI, for both research and industry.
4️⃣ Cleaning, organizing, chores: OpenAI targets everyday life
The summary : AI is no longer confined to screens. OpenAI claims it is about to ease everyday household tasks. According to Aaron “Ronnie” Chatterji, the company’s chief economist, artificial intelligence will save a significant amount of time at home, a benefit still absent from traditional economic indicators.
Details :
AI enters the home: On January 8, in the Financial Times, Aaron “Ronnie” Chatterji explained that AI will take over long, repetitive household tasks to free up daily time.
Productivity that’s hard to measure: According to OpenAI, domestic help, household organization, and family management remain largely invisible in traditional economic statistics.
ChatGPT, parent edition: In December, Sam Altman said on The Tonight Show that he could not imagine learning how to care for a newborn without ChatGPT.
Anthropic gets its hands dirty: A study conducted by Anthropic shows that using its chatbot Claude for meal preparation speeds up the process by 75%.
Redefining human value: Chatterji wants to develop new indicators that include domestic labor, child-rearing, and human collaboration.
Massive usage already visible: OpenAI reports that 40 million people use ChatGPT daily for health-related questions. Out of 800 million weekly users, 200 million address these topics at least once a week, accounting for more than 5% of all global messages.
Why it's important : AI doesn’t just threaten jobs, it also reshapes what we consider useful work. By valuing time freed at home, OpenAI proposes a shift in perspective: productivity no longer starts at the office, it begins in the living room.
5️⃣ Google wants AI to do your grocery shopping with Carrefour
The summary : Unveiled on January 11, 2026, in New York, Google’s Universal Commerce Protocol opens the era of agentic commerce. This standard makes it possible to purchase directly through a conversational AI, without leaving the interface. Carrefour is among the first adopters, alongside Visa and Zalando. A clear counter-move to ChatGPT’s shopping features launched three months earlier.
Details :
Shopping through conversation: Agentic commerce relies on a natural dialogue with an AI. The user requests a product, the assistant finds it, selects it, and completes the purchase within the same conversational flow.
A universal protocol: With the Universal Commerce Protocol (UCP), Google replaces custom integrations with a shared language. A single connection makes merchants accessible across all compatible AI assistants.
The full purchase journey covered : Search, add to cart, payment, and order tracking are integrated into this open-source protocol, while merchants retain control over pricing, promotions, and inventory.
A global alliance : Shopify, Etsy, Wayfair, Target, and Walmart took part in the protocol’s creation. More than 20 additional groups followed, including Carrefour, Zalando, Best Buy, Macy’s, and The Home Depot.
High-security payments : Each transaction relies on cryptographic proof of user consent. Visa, Mastercard, American Express, Adyen, Stripe, and PayPal support the UCP.
Carrefour as a European pioneer : Already using Google AI through its “Hopla+” agent, Carrefour is now integrating with Google Search and the Gemini app. Emmanuel Grenier, Chief Digital Transformation Officer, claims a strategic head start.
Why it's important : Shopping is moving away from clicks and into conversation. By standardizing AI-driven purchasing, Google is redefining e-commerce, and Carrefour is positioning itself early within this new buying reflex.
❤️ Tool of the Week : Email enters assisted-pilot mode with Gmail + Gemini
Google is turning Gmail into an intelligent assistant, with Gemini now at the core of the inbox. Natural-language search, automatic summaries, writing assistance, message prioritization… email is no longer read, it’s consulted.
What is it for?
Gmail no longer just displays messages. It understands, summarizes, and prioritizes them for you.
Find information without digging
You ask a question (“What was the plumber’s quote?”), Gemini scans your emails and answers directly, with key information and cited sources.
Automatically summarize endless threads
Long conversations turn into clear summaries. No more endless scrolling.Improve emails before sending
An AI proofreading tool suggests clearer, smoother reformulations, far beyond basic spellchecking.Let AI sort your inbox
The upcoming “AI Inbox” highlights priority messages and summarizes the rest. Your attention becomes an optimized resource.
How to use it?
The new features are rolling out gradually for Google AI Pro and Ultra subscribers. Some previously premium-only AI tools are also being deployed more widely. The AI Inbox itself is still being tested with a limited group of users.
💙 Video of the Week : Atlas does a backflip… and recovers
At CES 2026, Boston Dynamics brought out its best crowd-pleaser: Atlas, its humanoid robot, capable of walking like a human… and attempting a backflip in front of an audience.
The move isn’t perfect. Atlas rotates cleanly in midair, lands on its feet, stumbles slightly, loses part of its gripper, then stabilizes itself in a fraction of a second. And that’s precisely where the video becomes interesting. Less for the flip itself than for the instant recovery, a clear demonstration of an impressive level of motor control and real-time computation.
This isn’t Atlas’s first backflip. The robot pulled off the same maneuver eight years ago, and more recently even did it wearing a Santa suit. But the message here is different. It’s no longer just about showing off a flashy stunt, but about demonstrating ultra-fast coordination across dozens of joints, capable of correcting an error in real time.
Of course, a backflip doesn’t say much about Atlas’s ability to work in a warehouse or factory. Boston Dynamics has already showcased that side elsewhere. This video serves a different purpose: a reminder that humanoid robots are no longer stiff, predictable, or fragile. They’re becoming dynamic, adaptive, and almost intuitive in their movements.
Do you think your job is threatened by AI ?
