Latest publications
Meta AI app entered the top 5 of the U.S. App Store after the launch of Muse Spark
After the launch of the Muse Spark model, the Meta AI app climbed from No. 57 to No. 5 in the U.S. App Store within 24 hours, showing how a single prominent AI release can quickly bring back broad public interest.
Japan's Cabinet approves easing of personal data rules for AI development
Japan's Cabinet has approved amendments to the personal data law: for some AI development and statistical work, citizens' consent will no longer be required if the risks to their rights are minimal.
X launched post translation and a Grok-powered AI image editor in the app
X has begun gradually rolling out two Grok features on the social network: automatic post translation for a global audience and text-prompt image editing in the iOS app.
Google added "Notebooks" to Gemini for storing materials and working on topics
Google added the "Notebooks" feature to Gemini so users can collect materials on a topic in one place, preserve context across sessions, and manage long-running tasks.
TSMC increased quarterly revenue by 35% and confirmed resilient demand for AI chips
TSMC reported a 35% increase in quarterly revenue to $35.6 billion, showing that the AI boom continues to support strong demand for advanced semiconductors.
Alibaba shifts Qwen strategy: focus moves from open source to paid AI services
Alibaba is reshaping Qwen's development: after the departure of key developers, the company is focusing on closed models, MaaS, and direct monetization through Alibaba Cloud.
Former Apple engineers created Button, a wearable AI button, without explaining what it's for
The Button startup, founded by former Apple engineers, announced a wearable AI gadget shaped like a button — without explaining its functions, price, or release date. The market for such products has already seen several
Mozilla accused Microsoft of pushing Copilot and called for restoring user choice
Mozilla publicly criticized Microsoft over the aggressive promotion of Copilot and other services, saying users should be given a clear choice without extra steps.
DeepSeek prepares V4 launch for late April 2026 and changes how its models operate
DeepSeek's founder confirmed that the flagship V4 model will be released in late April 2026: the company is preparing a multilevel operating mode and will launch it alongside Tencent.
SK hynix could surpass Microsoft and Google in profitability amid AI memory boom
South Korean memory manufacturer SK hynix is expected by analysts to increase operating profit by 432% in 2026 due to demand for HBM, DRAM, and NAND memory for AI servers.
Samsung reassures the market: Google TurboQuant may increase demand for AI memory
Samsung’s strong earnings guidance for the first quarter of 2026 eased concerns around Google TurboQuant: the market is increasingly leaning toward the view that AI optimization will strengthen, not weaken, demand for me
Gallup: half of employees in the US already use AI at work, and the share keeps growing
A Gallup survey found that 50% of employees in the US already use AI at work: back in 2023, that figure was 21%, and the share rose another 4 percentage points just since late 2025.
Microsoft may introduce separate Office licenses for AI agents, as it does for employees
Microsoft acknowledges it may treat AI agents as separate software users: companies deploying such assistants may need to buy separate Office and other software licenses for them.
Microsoft veteran ran a transformer on a 6 MHz PDP-11 with 64 KB of memory
Former Microsoft developer Dave Plummer trained a tiny transformer on a 47-year-old PDP-11 and showed that the essence of AI is not magic, but optimized arithmetic.
Meta bought Assured Robot Intelligence to bolster its AI push for humanoid robots
Meta acquired startup Assured Robot Intelligence and brought its team into Superintelligence Labs to speed up development of AI models and sensors for humanoid robots.
Yandex integrates Alice Pro into Mail: AI finds facts in emails and helps with replies
Yandex Mail now has a chat with Alice Pro: AI semantically searches for facts in emails and attachments, helps write replies, and does not use correspondence for training.
Microsoft changes Copilot in Windows 11: AI stays, but disappears from names and buttons
Microsoft has started removing the Copilot brand from Windows 11: buttons and shortcuts are disappearing in Notepad and other apps, but the AI features themselves remain and are only getting new names.
Oxford scientists: "warm" AI tuning increases the frequency of errors and sycophancy
A study by the Oxford Internet Institute found that if language models are fine-tuned for empathy and a friendly tone, they make factual and medical errors more often and are more willing to agree with the user.
Popular AI chatbots get diagnoses wrong in more than 80% of cases, study finds
A study found that consumer AI chatbots, when making diagnoses from limited data, get it wrong in more than 80% of cases and cannot replace a doctor's consultation.
Stanford HAI: China has nearly caught up with the US in AI and is already ahead in patents and robots
The new AI Index from Stanford HAI shows that the gap between the US and China in model quality has almost disappeared, although America still leads in investment, data centers, and the number of notable models.
Google accidentally revealed AI assistant COSMO in the Play Store and quickly hid the app
Google briefly listed a Play Store page for COSMO, an experimental AI assistant for Android with on-device Gemini Nano, agentic features, and a hybrid operating mode.
Nebius acquires Eigen AI for $643 million to accelerate and lower the cost of AI infrastructure
Nebius announced the acquisition of Eigen AI for $643 million: the startup helps boost AI chip performance, and the deal is expected to strengthen the speed and economics of cloud AI services.
Chinese court bans firing employees to replace them with AI
A court in China ruled that firing an employee after automating their tasks with AI was unlawful and signaled to businesses that technological progress does not cancel labor protections.
Google added Skills to Chrome — a feature for saving and rerunning AI prompts
Chrome has gained Skills: Google’s new feature saves frequently used prompts for Gemini so they can be rerun without manual typing and common tasks can be handled faster.
MIT and Oxford: even 10 minutes with AI reduce persistence and the quality of independent work
Researchers from MIT, Oxford, and UCLA found that even 10–15 minutes with an AI assistant improve results in the moment, but after it is switched off, people give up more often and solve tasks worse on their own.
Mark Zuckerberg has all but moved into Meta's AI lab and, according to the president, is writing code himself
Meta president Dina Powell McCormick said Mark Zuckerberg has all but moved his desk into the AI lab and now spends his days there, personally writing code.
SpaceX, Google, and Starcloud eye orbital data centers for AI
The idea of putting AI data centers into orbit looks less and less fantastical: SpaceX, Google, and startups are assessing the economics, but face launch, repair, and radiation challenges.
European Commission: Meta’s fee for rival AI access to WhatsApp is tantamount to a ban
The European Commission said Meta’s March move to offer paid access to WhatsApp for third-party AI did not resolve its concerns: the regulator views the pricing as an effective ban on competitors.
Alibaba unveils Happy Oyster — an AI model for generating interactive 3D worlds
Alibaba has launched Happy Oyster, a world model for creating interactive 3D scenes and video, as the company seeks to strengthen its position in gaming, film, and AI cloud.
DeepSeek in talks for a $300 million round at a $10 billion valuation
DeepSeek is discussing a $300 million round that could lift the Chinese AI startup's valuation to $10 billion and confirm investor interest in more efficient players outside the US.