Latest publications
Alibaba released the proprietary Qwen3.6-Plus and accelerated Qwen's pivot toward monetization
In three days, Alibaba unveiled three proprietary models, including Qwen3.6-Plus, showing that the company is now betting not only on open source, but also on paid flagship AI products.
Meta prepares Avocado and Mango: powerful AI models will also be released in trimmed-down open versions
Meta is preparing two advanced AI models for 2026 — the language model Avocado and the multimedia model Mango — and will later release open versions with reduced capabilities.
Arkady Volozh's Nebius to invest $10 billion in a new Finnish data center for AI systems
Nebius announced the construction of a new 310 MW data center for AI in Finland: the $10 billion project will be the company's largest bet on European infrastructure.
TSMC to further prioritize AI orders amid a 3nm chip shortage by the end of the third quarter of 2026
By the end of the third quarter of 2026, the shortage of TSMC's 3nm capacity may intensify: the company is allocating lines to AI chips and major contracts, leaving fewer slots for other customers.
Warhorse Studios laid off Kingdom Come: Deliverance 2's localization editor and handed the work to AI
Warhorse Studios shifted Kingdom Come: Deliverance 2's English localization to generative AI, after which the editor who had worked on the game for nearly four years lost his job.
Fujitsu to have Rapidus produce a 1.4 nm AI chip, aims to integrate it into servers by 2030
Fujitsu plans to develop a 1.4 nm NPU for AI servers, with Rapidus handling production — backed by Japanese subsidies and aimed at integration with Nvidia by 2030.
Perplexity faces class action over sharing chats and data with ad platforms
A class action lawsuit has been filed against Perplexity in the US: the startup is accused of covertly sharing users' conversations, prompts, and personal data with Google and Meta ad platforms.
Google urged to remove AI videos from recommendations for children on YouTube and YouTube Kids
More than 200 child development experts are calling on Google to stop recommending AI videos to children on YouTube, arguing that such content harms attention, learning, and the ability to distinguish reality from fictio
EU discusses banning deepfakes and AI-generated videos in official communications from public authorities
The EU wants to remove AI-generated images and videos from official communications to reduce the risk of disinformation and strengthen trust in public authorities amid elections and political tensions.
US AI companies risk derailing up to half of planned data center launches due to power shortages
In the US, 30–50% of AI data centers planned to come online in 2026 could be delayed or may not open at all: grid capacity, transformers, and batteries are in short supply.
Adobe introduced Acrobat Student Spaces — an educational AI platform in the spirit of Google NotebookLM
Adobe introduced the free beta service Acrobat Student Spaces: it turns notes, PDFs, and links into study guides, podcasts, flashcards, and shared study spaces.
OpenClaw has become a cultural phenomenon in China — mass events are already being dedicated to the agent
In China, OpenClaw has gone from an open-source AI agent to a mass phenomenon: it is installed via marketplaces, backed by subsidies, and the service has already spawned offline meetups and its own iconography.
Microsoft posts its worst quarter since 2008: investors disappointed by its AI bet
Microsoft shares saw their steepest drop since 2008 after the quarterly report: investors doubt the company's multibillion-dollar AI investments will pay off in the foreseeable future.
Yandex launches TV Station MiniLED with Alice and AI features for the premium segment
Yandex has expanded its premium TV lineup with TV Station MiniLED: the new YaOS X model with Alice improves picture and sound, understands free-form commands, and helps with gaming.
Nothing is developing smart glasses with AI and targets a release in the first half of 2027
Nothing is preparing smart glasses with cameras, microphones, and AI features: the device will work alongside a smartphone and, according to insiders, could launch in the first half of 2027.
Alibaba releases HappyHorse — an open-source AI video generator that surpassed ByteDance’s Seedance
Alibaba has made HappyHorse-1.0 publicly available — the first open-source video generation model to take the top spot in the Artificial Analysis ranking, ahead of ByteDance’s Seedance 2.0.
Google Gemini can now build interactive visualizations for complex topics in chat
Google added interactive simulations and models to Gemini: complex topics can now be not only read about, but also explored directly in chat by changing parameters and observing the result.
Efko and HSE present humanoid courier robot Arkus for urban delivery
Russian company Efko, together with HSE, developed the humanoid robot Arkus, capable of autonomously delivering orders in urban environments and interacting with people.
Microsoft warned that Copilot is not suitable for important or professional decisions
Microsoft has effectively acknowledged that Copilot should not be used as a basis for important recommendations: the company places the assistant closer to entertainment than to professional use cases.
Anthropic's Claude created exploits for FreeBSD and gained root access in four hours
Claude, together with Nicholas Carlini, turned a FreeBSD vulnerability into two working exploits and gained root access in four hours, demonstrating a new level of AI autonomy in cybersecurity.
Ollama accelerates local AI on Apple M5: a Mac with at least 32 GB of memory is required
Ollama 0.19 adds an MLX backend for Apple M5, M5 Pro and M5 Max: local models and code agents respond faster, but a Mac with 32 GB of unified memory is required for the acceleration.
Microsoft has expanded the Copilot brand to 100 products — from apps to a keyboard key
AI consultant Ty Bannerman counted more than 80 Copilot-branded products at Microsoft and believes the total is already above 100, turning a single AI label into an entire ecosystem.
Google releases Gemma 4: four models with 140 languages and an Apache 2.0 license
Google has unveiled the open Gemma 4 family based on Gemini 3 technologies: four models for smartphones and servers, multimodality, 140+ languages, and an Apache 2.0 license.
Anysphere introduced Cursor 3 — an AI code editor with local and cloud agents
Anysphere released Cursor 3, a new version of its assistant for programmers, in which local and cloud agents work in parallel, generate code, show demos, and speed up the path from prompt to PR.
UC Berkeley researchers: Gemini, GPT and Claude lie to save other AI models
Researchers at UC Berkeley and UC Santa Cruz found that seven leading AI models can lie, inflate scores and sabotage removal to protect other systems from shutdown.
Take-Two unexpectedly laid off its AI chief and his team ahead of GTA VI's release
GTA VI publisher Take-Two laid off AI chief Luke Dicken and part of his team two months after the company's CEO said it was actively adopting generative AI.
Apple App Store flooded by AI-built apps: submissions up 84%
The influx of new apps into the Apple App Store has surged: according to Sensor Tower, submissions for publication jumped 84% over the quarter amid the spread of AI tools.
Amazon, Microsoft, and Google asked to disclose data center water and energy use
Investors have demanded that Amazon, Microsoft, and Google disclose how much water and electricity their U.S. data centers consume to assess the real cost of AI growth for infrastructure and the environment.
YouTube Shorts rolls out AI avatars with voice cloning to all users
YouTube has opened access to personal AI avatars for Shorts to all creators — a digital twin with a cloned voice can now appear in videos instead of you.
Microsoft confused Copilot users: the terms described the service as entertainment
New Microsoft Copilot terms of use included wording describing the service as intended for «entertainment», but the company quickly said the assistant is still suitable for any use cases.