The latest IT news from the tech world
Follow with us the freshest news from the world of IT, software, and hardware. We will discuss products from companies such as Google, Microsoft, Facebook, OpenAI, and others. We will also present IT news from hardware giants Intel, Nvidia, AMD, as well as technology updates from robotics, smartphones, and gaming.
In the article you will learn:
November 2025
YouTube adds a timer to Shorts
After six months of testing, YouTube is making a timer available for short videos (known as Shorts), which is designed to help users control the time spent watching short-form content. After setting a daily limit and reaching it, the app will display a notification and temporarily stop scrolling, which the user can cancel at any time. Videos will continue to appear in other parts of the app, so this is not a hard restriction but rather a digital reminder of the time spent with Shorts.
The update expands YouTube’s existing digital well-being tools, which include break reminders or bedtime settings. This time, however, it focuses exclusively on Shorts – a format that most supports uncontrolled content consumption. The timer also becomes part of parental controls, where the notification cannot be bypassed. YouTube is thus catching up with competing platforms that have been offering similar limits for a longer time and is responding to the increasing demand for better control over time spent on social networks.
Source: https://www.androidauthority.com/youtube-shorts-time-limits-3609271/
Claude Skills – capabilities that speed up AI and reduce its costs
Claude AI is introducing a new feature called Claude Skills – capabilities tailored for specific tasks. Users can create their own Skills that contain precisely defined procedures, commands, or even scripts. The AI will no longer have to interpret the instructions from scratch each time; instead, it activates a prepared module designed exactly for the given task.
This approach provides higher speed, accuracy, and lower costs, especially when used via API. While today’s AI models often switch between simple and deep text processing, Skills eliminate less efficient methods. The AI performs only what the module prescribes. This is particularly advantageous for automating routine processes, processing spreadsheets, creating presentations, or working with graphics adapted to corporate standards.
Claude Skills are therefore an extension of standard conversational models: the more precisely the user defines the module, the less they have to explain what they want to achieve with each prompt.
Source: https://claude.com/blog/skills
Wikipedia warns that AI bots are stealing traffic and reducing the number of human visitors
Wikipedia states in its official blog that in the monitored period, it recorded approximately an 8-percent year-on-year decrease in visits by real users. At the same time, in May and June 2025, more bot-generated requests appeared. According to Marshall Miller, this is partly caused by search engines and AI tools that provide ready-made summaries to users. People often get their answers immediately and therefore have no reason to click through to the original article.
This poses a risk to the encyclopedia itself. Fewer real visitors also mean fewer volunteer editors and less quality control of the content. Miller recommends verifying information by clicking on the original sources. AI can provide a quick overview, but for better understanding of a topic and to support the authors, it is always advisable to read the full article.
Source: https://diff.wikimedia.org/2025/10/17/new-user-trends-on-wikipedia/
Windows 11 tightens security settings for automatic file previews
Microsoft is tightening security settings and changing how File Explorer in Windows 11 handles risky documents. Previews will no longer appear for files marked with the Mark of the Web attribute. This attribute is automatically added to all files downloaded from the internet, email attachments, or materials from shared storage that may contain malicious code.
Following the example of Microsoft Office, which has for years been opening internet documents in read-only mode, File Explorer will now take a similar approach. If a file carries the MotW attribute, its preview will not load until the user manually unblocks it. The change will take effect after October 14, 2025.
The preview can be restored easily – simply select the file, open its properties (Alt + Enter), and confirm the Unblock option.
Source: https://support.microsoft.com/en-us/topic/file-explorer-automatically-disables-the-preview-feature-for-files-downloaded-from-the-internet-56d55920-6187-4aae-a4f6-102454ef61fb
ChatGPT now allows sharing of projects
ChatGPT is adding a useful new feature – project sharing. Projects serve to organize conversations into thematic groups; they have their own name, icon, and, most importantly, instructions that define the response style (language, format, sources, processing method).
The update introduces the ability to share an entire project with other users, both on the web and in the mobile app. Collaborators will see the full conversation history, but ChatGPT does not store this information in user memory. This feature is practical for team collaboration, vacation planning, or short-term tasks. Unlike the original sharing of individual conversations, the new system is fully cooperative. It allows editing the project, moving existing chats into it, and keeping topics neatly organized in the left panel.
Source: https://help.openai.com/en/articles/10169521-projects-in-chatgpt
Microsoft revives Clippy, the iconic paperclip
After years, Microsoft is bringing back the iconic Office paperclip, but in an entirely new form. The new AI assistant Mico, integrated into Copilot, builds on the famous Clippy from Office 97 – 2003. This time, however, it is not just a nostalgic gesture but an attempt to create a friendly digital assistant with modern capabilities.
Mico will change shapes and colors, and its animated appearance resembles a soft, moldable piece of chewing gum rather than the original paperclip. Microsoft wants it to feel more human and less intrusive than Clippy, which was previously criticized for annoying prompts. Users will be able to choose from eight voices (both female and male), and it will even be possible to switch the visual design back to the classic Clippy.
The features, which include group conversations in Copilot and the assistant’s new appearance, are currently available only in the USA. They will expand to the United Kingdom and Canada in the coming weeks. There is currently no information about availability in Slovakia.
Source: https://www.theregister.com/2025/10/24/microsoft_clippy_copilot_update/
Musk introduces Grokipedia – his own alternative to Wikipedia
Elon Musk is launching Grokipedia, a new free encyclopedia designed to compete with Wikipedia. The project is currently in version 0.1, yet it already contains more than 885,000 articles. The homepage resembles an AI tool rather than a classic encyclopedia; the minimalist interface offers only a search field.
Grokipedia relies on content from Wikipedia, which is licensed under Creative Commons, and then AI restructures and refines it into a clearer, more organized format. However, you will not yet find images in it, nor can you expect Slovak language support. Since the texts are generated or edited by artificial intelligence, there is a risk of inaccuracies or hallucinations.
Musk claims he is removing propaganda from Wikipedia and creating a more objective source. However, the automation of article processing is not yet fully transparent. The project is in an early stage, and its quality will depend on further development, which may take months or even years.
Source: https://grokipedia.com/
Google Translate adds a toggle between speed and accuracy
Google is preparing a major update to Translate that will allow users to choose between a fast and an accurate translation. A new toggle at the top of the app will offer two modes: Advanced – maximum accuracy for complex texts, and Fast – an instant translation for shorter texts, suitable for signs or menus.
In addition, Google is also testing new AI features: Understand will explain why the text was translated in a certain way. Ask will allow users to request an alternative tone, regional variant, or specific adjustment of the translation.
The new features are already available in iOS 26.2 Beta 1, with Android support coming later. Google has not yet specified whether the more accurate model will be paid, but it is not ruled out for demanding translations.
Source: https://www.phonearena.com/news/google-translate-users-will-make-important-decision-before-employing-app_id175473
Firefox strengthens anti-tracking protection
Firefox version 145 brings a significant improvement to protection against so-called fingerprint tracking – a technique that collects settings and system data through JavaScript and creates unique “digital fingerprints” of users. This method can identify and track users even without cookies and affects private (incognito) windows as well.
Mozilla has been blocking known trackers since 2020 and strengthened its protection in 2021, including methods related to graphics cards. Now it is introducing another level of fingerprinting defense. If you care about privacy, consider Firefox or its derivatives (LibreWolf, Waterfox) instead of Chrome. Mozilla also confirmed continued support for Windows 10.
Source: https://blog.mozilla.org/en/firefox/fingerprinting-protections/
ChatGPT 5.1 brings more personal conversation and new Instant and Thinking models
OpenAI has released the updated GPT-5.1 model for paid accounts (Pro, Plus, Go, Business). The update introduces two modes: Instant and Thinking. Instant is intended to be faster, more natural in conversation, and better at following user instructions, while Thinking offers more thoughtful responses with the option of longer or shorter reasoning.
The option to display dual output, known from GPT-5, remains available. The user will see two responses side by side, and by selecting the preferred one, the model gradually tunes its communication style. Developers will appreciate the faster responsiveness of Instant and the more powerful Thinking mode for demanding tasks, although the latter may require more processing time. The new models will be available via API within a few days.
Older versions will continue to function for another three months, giving users enough time to compare results and gradually transition to the new, more personal mode of interaction.
Source: https://openai.com/index/gpt-5-1/
NAND prices are rising sharply
Phison CEO Khein-Seng Pua has confirmed dramatic developments in the market. NAND memory prices have more than doubled in the last six months, and they are expected to continue rising. The reason is extreme demand driven by the boom in artificial intelligence. Data centers are switching from HDDs to SSDs on a large scale to avoid long delivery times.
The entire NAND production capacity for 2026 is already sold out, meaning the market will face NAND shortages at least until the end of 2027, until new factories come online. Prices are rising across the entire segment: for example, a 1 TB TLC chip has increased from USD 4.80 (July) to USD 10.70, and even older MLC memories have doubled in price.
Consumers will feel the consequences most, as manufacturers prioritize enterprise customers with larger budgets. According to Phison, the rate of price growth continues to accelerate, and the market is facing a long-term imbalance between supply and demand caused by the rapid rise of AI and insufficient investment by manufacturers in capacity expansion.
Source: https://www.tomshardware.com/pc-components/ssds/phison-ceo-confirms-nand-prices-have-more-than-doubled-and-will-continue-to-rise-all-2026-production-already-sold-out-ssds-facing-pricing-apocalypse-throughout-2027
Visual Studio 2026 is coming
Microsoft has introduced Visual Studio 2026, which brings a modernized Fluent UI interface, faster project loading, and overall better performance even with large solutions. The new version can be installed alongside Visual Studio 2022, making the transition risk-free.
The IDE now supports .NET 10, C# 14, and C++ 23/26, while the improved MSVC toolchain and new Profiler Agent enable more detailed memory and CPU analysis. This makes it easier to identify performance bottlenecks.
A key update is that the Community and Professional editions also receive AI-based Code Coverage, previously available only in the Enterprise version. According to Microsoft, AI is used here for real testing improvements, not just marketing.
Visual Studio 2026 works best on Windows 11 and modern multi-core hardware with 64 GB RAM, but it is also compatible with Windows 10 (22H2). Overall, it represents the biggest leap forward in recent years. The combination of a more modern interface, higher performance, and AI integration moves the IDE another generation ahead.
Source: https://www.techspot.com/downloads/7808-visual-studio-2026.html
Unreal Engine 5.7 – MegaLights, Nanite Foliage, and AI assistant
Epic Games is introducing Unreal Engine 5.7, which brings one of the biggest innovations in recent years. The main new feature is MegaLights technology, which has entered beta. It brings more realistic directional lighting, better transparency of materials, and more accurate shadows for particles and hair. Developers can expect better out-of-the-box performance, less noise, and significantly reduced need for manual optimization.
Another major innovation is the experimental Nanite Foliage technology, which finally allows efficient vegetation rendering. Thanks to the Nanite Voxels system, the engine can render millions of tiny overlapping details (leaves, needles, tree crowns) without performance drops and without the need to manually rework distant objects.
Version 5.7 also introduces Substrate – a system for creating complex, layered, physically accurate materials that pushes visual fidelity even further. A bonus is a built-in AI assistant directly in the editor, which can answer questions or generate C++ code. Improvements to MetaHuman tools and animations are also included.
Source: https://wccftech.com/unreal-engine-5-7-out-now-with-nanite-foliage-and-megalights-powered-stunning-dynamic-shadow-casting-lights/
OpenAI lost a court case in Germany regarding training on song lyrics
OpenAI is facing its first significant legal defeat in Europe for using copyright-protected song lyrics to train its language models. The German organization GEMA, which represents more than 100,000 composers, sued the company over 9 songs that appeared in ChatGPT’s outputs. The court ruled that OpenAI used these works without a license and therefore must pay not only for their use during training but also for the generated outputs during inference.
The company argued that the models do not store lyrics but only learn from them, and that users are responsible for any output they generate. The court rejected this. This ruling may fundamentally influence future legal debates on AI in Europe and suggests that LLM developers will no longer be able to avoid licensing the data they use.
Source: https://techxplore.com/news/2025-11-german-music-body-copyright-case.html
XPeng Iron – a humanoid robot with realistic movement and advanced AI
At AI Day 2025, XPeng unveiled the new generation of its humanoid robot Iron, whose appearance and movements closely resemble a human. It features synthetic skin with touch sensors, 22 degrees of freedom, and a surface structure imitating muscles.
The robot uses three XPeng Turing AI chips with a total performance of 2250 TOPS, running advanced multimodal models VLA 2.0, VLM, and VLT. These provide conversational abilities, autonomous planning, and task execution. The technologies are trained on massive datasets, and some of them are expected to be used by Volkswagen as well.
In practice, Iron will initially appear primarily in marketing or assistant roles, as its high price prevents immediate deployment in homes or industry. Production will begin in April 2026 and should transition to mass production by the end of the year. XPeng is also preparing an SDK for developers.
Source: https://cnevpost.com/2025/11/06/xpengs-iron-humanoid-robot-looks-so-real/
Gemini 3 introduces a graphical mode
Google introduced Gemini 3 – a multimodal model focused on accuracy, vibe coding, and rich visual presentation of responses. The model offers modes such as Deep Research, Canvas, and an agent mode, aiming to be more direct and factual without unnecessary filler.
The most interesting novelty is the graphical mode – an experimental feature in the Tools menu that combines text, images, tables, and interactive elements into a single response (processing usually takes more than 10 seconds). The result feels more like a small web page rather than the plain text we are used to from ChatGPT.
Source: https://blog.google/products/gemini/gemini-3-gemini-app/
A trivial Cloudflare error took down a quarter of the Internet for hours
On November 1, 2025, the Internet experienced the largest Cloudflare outage in six years. Millions of websites were inaccessible – from banks and e-shops to social networks and gaming platforms. Cloudflare is a key, though invisible, piece of infrastructure for users, accelerating and securing roughly 20 to 25% of the entire web.
The problem was not caused by a DDoS attack but by a surprisingly simple human error. A permissions change in one database propagated a faulty function file across the entire network. The file exceeded the allowed size, overloaded proxy systems, and triggered a massive wave of HTTP 5xx errors. Services were disrupted from 12:20 to 15:30, with full recovery occurring shortly after 17:00.
Cloudflare promises that such an incident will not happen again. It is introducing stricter limits for configuration files, more global “kill switches,” and an extensive review of proxy modules.
Source: https://blog.cloudflare.com/18-november-2025-outage/
October 2025
MediaTek Dimensity 9500 – a new 3nm chip delivers record performance and efficiency for Android
MediaTek introduced its most powerful chipset, the Dimensity 9500, the first flagship processor for 2026. The new chip uses TSMC’s 3nm manufacturing process and combines high CPU, GPU, and AI performance with a focus on efficiency and cooling. Thanks to Dimensity Scheduling 2nd gen and Super zRAM, it can run up to 20 apps without performance loss and achieves a response time of just 42 ms.
The chip includes a 9th-generation NPU that doubles AI performance, enables 4K image generation directly on the device, and improves efficiency when handling models like Gemini. The Arm G1-Ultra GPU provides 119% faster ray tracing and supports up to 120 fps, ensuring a console-level gaming experience.
Support includes cameras up to 320 MP, 5G with AI optimizations, Wi-Fi 7, improved Bluetooth, and more precise GNSS positioning. The Dimensity 9500 opens a new generation of mobile chips and will soon compete with the Snapdragon 8 Elite Gen 5 and Exynos 2600 for the title of the most powerful Android processor.
Source: https://www.mediatek.com/products/smartphones/mediatek-dimensity-9500
Snapdragon 8 Elite Gen 5 – the fastest mobile system-on-chip in the world
Qualcomm launched the Snapdragon 8 Elite Gen 5 chipset, which has become the most powerful mobile processor to date. It is manufactured using TSMC’s 3nm process and uses the 3rd generation of Qualcomm Oryon cores, delivering 20% higher CPU performance and 35% better energy efficiency. The new Adreno GPU boosts graphics performance by 23% while reducing power consumption by a fifth.
The chip introduces a built-in AI assistant capable of analyzing user behavior, recognizing images and sound, and performing tasks autonomously – without sending data off the device. It also offers the Advanced Professional Video (APV) codec, enabling professional video recording and post-production directly on a smartphone. Snapdragon 8 Elite Gen 5 supports the X85 5G modem, an improved Hexagon NPU with a 37% performance increase, and AI optimizations for connectivity.
It will appear in flagship models from brands such as Samsung, Xiaomi, OnePlus, Sony, OPPO, vivo, Honor, and ROG, with the first devices featuring this chip expected to be unveiled in the coming days.
Source: https://www.qualcomm.com/news/releases/2025/09/snapdragon-8-elite-gen-5–the-world-s-fastest-mobile-system-on-a
LinkedIn will begin training AI on user data
LinkedIn announced the return of a project in which it will train its AI models on user data. After criticism in 2024, when these activities were temporarily halted, the social network now openly admits it will use public profile information, such as statuses and résumés. Private messages will not be included in training. The changes take effect on November 3, 2025, and LinkedIn will simultaneously begin sharing more data with Microsoft, its parent company, for targeted advertising.
Users will be able to opt out of providing this data, similar to the existing option for linking LinkedIn with Microsoft 365. The adjustments mainly concern countries in the European Economic Area, where GDPR applies. Although this regulation protects user privacy, it allows tech companies to collect certain data if they demonstrate necessity, similar to cookies essential for proper website functionality.
Source: https://hostvix.com/linkedin-will-soon-train-ai-models-with-data-from-european-users/
ChatGPT Pulse – a morning AI briefing
OpenAI is introducing ChatGPT Pulse, a feature designed to make ChatGPT part of the user’s daily morning routine. It provides an intelligent daily briefing summarizing topics the user recently discussed with the chatbot, enhanced with news, events, and tips for the upcoming day. Unlike previous usage patterns, Pulse is designed to act proactively, initiating conversations and offering content without requiring manual activation.
The tool can connect to Gmail and Google Calendar, informing users about important emails, meetings, or tasks. The briefing is prepared overnight and is ready in the app upon waking. For now, Pulse is available only to ChatGPT Pro subscribers, with future rollout planned for ChatGPT Plus and the free tier.
Source: https://openai.com/index/introducing-chatgpt-pulse/
Windows 10 will receive one additional free year of support in the EU
Microsoft will extend Windows 10 support by one year for free, but only for users in the European Economic Area (EEA). The system was originally slated to reach end of support on October 14, 2025, but after pressure from the organization Euroconsumers, Microsoft will extend it until October 13, 2026. The decision follows a request for longer support to reduce electronic waste and enable safe use of older devices.
Outside the EU, the extension remains paid: users must have a Microsoft cloud account and pay $30 or redeem 1000 Microsoft Rewards points. Euroconsumers notes that even a three-year extension after the end of the lifecycle is insufficient, as Windows 10 is still used on more than 850 million devices.
Compared to previous versions, this is the shortest support window: Windows 10 has only 4 years after the release of Windows 11, while Windows 7 and XP had 7 to 8 years of support following their successors.
Source: https://www.euroconsumers.org/microsoft-security-windows-10-planned-obsolescence/
Claude 4.5 surpassed ChatGPT-5 in programming tests
Anthropic claims it is the best model for developers in the world. According to internal tests, the new Claude Sonnet 4.5 model has become the best AI model for programming, surpassing even OpenAI’s ChatGPT-5. Although this assessment comes from the company itself, the results indicate substantial progress. One of the key innovations is the introduction of checkpoints during code creation, allowing developers to return to earlier versions of a project without the chaos caused by multiple changes. The model thus improves workflow clarity and reduces the risk of errors.
According to Sean Ward, CEO of iGent AI, Claude 4.5 can autonomously code for more than 30 hours straight, while the previous Opus 4 model managed only about 7 hours.
Premium users also gain access to a Chrome extension that enables interactive work with web-based spreadsheets, documents, and presentations directly in the browser. Additionally, Anthropic is introducing the Agent SDK, allowing Claude to connect coding with other applications, work with local files, and automate tasks in email or calendars.
Source: https://www.anthropic.com/news/claude-sonnet-4-5
Sora 2 brings more realistic videos and improved object physics
OpenAI has introduced Sora 2 – an enhanced model for creating realistic videos that significantly improves the physical behavior of objects as well as sound quality. The main new feature is the ability to insert your own face directly into the generated video, including vertical format for TikTok. The system can link video creation with the user’s ChatGPT conversation history, location, or previous activity, enabling more personalized content. However, concerns have been raised regarding data security, as uploaded faces may be shared between users. OpenAI is therefore adding parental controls to protect younger creators.
The first demos look very realistic. Sora 2 convincingly simulates motion physics (for example, a ball bouncing off various surfaces). Critics, however, point out that the videos still lack the human touch that can be felt in real footage. It is also facing competition from Google’s Veo 3 and Nano Banana, which are likewise striving for the highest possible level of visual realism. Which tool will ultimately prevail will only become clear through practice and real-world use.
Source: https://openai.com/index/sora-2/
YouTube changes its look after ten years
YouTube has undergone its biggest visual update in the past decade. The new design is gradually rolling out to smartphones, computers, and smart TVs. The aim is to unify the user experience and make the interface clearer, more modern, and more interactive.
The most striking change is the redesigned video player, which now takes up less space and is less distracting while watching. The controls have rounded shapes in the Material Design 3 style and a translucent background, making them appear lighter. On smart TVs, video information has been moved to the top left corner, while the mobile app groups controls into a compact capsule on the left side.
Google has also improved animations and responsiveness within the Material Expressive concept, making interactions feel smoother. New interactive effects appear when liking, saving videos, or commenting, and for example, music videos display a note or sports clips show an in-game effect. The redesign also includes clearer comment threads and aims to make YouTube “feel as dynamic as the creators who make content on it.” The update is already available in Slovakia.
Source: https://support.google.com/youtube/thread/380540176
Xiaomi and MediaTek bring GPS with millimeter accuracy
Xiaomi, MediaTek, and China Telecom have introduced a new localization system that, thanks to RTK (Real-Time Kinematic) technology, can determine position with millimeter accuracy. This represents a significant leap compared to standard GPS, which has an error margin of several meters.
RTK works by comparing satellite data with information from a fixed base station, eliminating signal errors. Until now, the technology was available only in specialized fields such as agriculture or robotics, but thanks to MediaTek chips, 5G connectivity, and Xiaomi software, it is reaching consumer devices for the first time.
Xiaomi plans to integrate the new RTK GPS module into upcoming smartphones, electric vehicles, and drones that will be part of its HyperConnect platform. This platform connects smart cities, IoT devices, and connected cars into a single ecosystem. Although the exact launch date has not yet been revealed, the system represents a new generation of navigation –fast, accurate, and ready for the era of autonomous transportation and smart cities.
Source: https://www.gizchina.com/xiaomi-phones/xiaomi-and-mediatek-unveil-millimeter-accurate-gps-a-new-era-for-smart-mobility
UFS 5.0: The new generation of flash storage delivers nearly double the speed and higher reliability
The JEDEC organization has officially introduced the new UFS 5.0 storage standard, which nearly doubles data transfer speeds compared to the previous version. The new memory modules achieve up to 10.8 GB/s, bringing them close to the speed of PCIe Gen 4 SSDs used in computers.
UFS 5.0 also brings technical improvements: signal equalization for more stable transfers, separate power delivery for lower noise, and inline hashing for increased data protection. It is also backward-compatible with UFS 4.x versions, making it easier for manufacturers to adopt the new technology. Production of the first devices with UFS 5.0 should begin soon. The new standard will deliver faster app loading, more efficient AI task processing, and overall performance comparable to modern SSDs.
Source: https://www.jedec.org/news/pressreleases/ufs-50-coming-jedec%C2%AE-sets-stage-next-leap-flash-storage
OpenAI will start designing its own AI chips
OpenAI has announced a strategic collaboration with the American company Broadcom, through which it will begin designing its own chips and accelerators tailored specifically for its AI models. This marks a major shift, with hardware now being adapted to software rather than the other way around, promising significantly higher performance and efficiency. The new compute farms will have power consumption measured in gigawatts.
OpenAI expects the first outcomes of the collaboration in the second half of 2026, with a dedicated 10-GW data center set to be built by 2029 – comparable to the combined electricity consumption of the Czech Republic and Slovakia. The company is thus expanding its partnerships with Nvidia and AMD, with the AMD agreement including an additional 6 GW of compute capacity. The goal is to meet the rising demand for artificial intelligence and push forward the limits of models such as ChatGPT.
Source: https://openai.com/index/openai-and-broadcom-announce-strategic-collaboration/
People will teach AI Grok to program games for 100 USD per hour
Elon Musk’s company xAI is launching a new project aimed at transforming its chatbot Grok into a game development assistant. To achieve this, the company needs high-quality training data and is therefore seeking video game experts who will teach the AI the principles of game design, mechanics, and development. These trainers will use xAI’s internal software to label data, write annotations, and explain game concepts so that Grok can better understand the game development process.
The pay for this work ranges from 45 to 100 USD per hour. Applicants should have experience in game design, IT, or interactive media, and ideally their own portfolio. The job is available at the company’s headquarters in Palo Alto, but xAI also allows fully remote collaboration, which is unusual given Musk’s long-standing preference for in-office work.
Source: https://www.techspot.com/news/109730-elon-musk-xai-pay-100-hour-video-game.html
Tesla shows Optimus robot performing kung-fu
Elon Musk has released a new video in which his humanoid robot Tesla Optimus performs the martial art of kung-fu. The short, roughly half-minute clip is meant to demonstrate progress in the robot’s development, especially in the areas of speed, movement fluidity, and the ability to maintain balance even during dynamic actions.
According to Musk, the scene was controlled by artificial intelligence, not by a human using remote control, although it is unclear to what extent the behavior was autonomous and to what extent it was a scripted demonstration. The video therefore appears more like an entertaining showcase of the robot’s potential rather than a practical demonstration of real capabilities. Nevertheless, it shows that the development of the Tesla Optimus humanoid robot continues at a rapid pace.
Source: https://interestingengineering.com/innovation/optimus-robot-performs-kung-fu-moves
AV2: A new generation video codec will deliver 30% smaller files at the same quality
The Alliance for Open Media has introduced the new AV2 format, which will replace the popular AV1. According to a presentation by Netflix expert Andrey Norkin, AV2 can maintain the same video quality at a 30% lower bitrate, which means smaller files and reduced requirements for internet connection speed.
The codec uses numerous new techniques, from advanced intra and inter prediction, through Chroma From Luma and Warp prediction, to expanded transformation and filter options. Thanks to this, it achieves higher efficiency but also higher computational demand. Playback without a hardware decoder will therefore require more powerful processors supporting AVX2 and AVX-512.
Google, Meta, Netflix, Apple, Tencent, Nvidia, Intel, Samsung, and other tech giants are collaborating on AV2’s development. The format is expected to be finalized as a standard by the end of 2025, with extended profiles and encoder optimizations expected in the following years.
Source: https://videocardz.com/newz/av2-video-codec-delivers-30-lower-bitrate-than-av1-final-spec-due-in-late-2025
Asus showcased the world’s most luxurious graphics card – 5 kg of real gold in the RTX 5090D
At the Bilibili World 2025 expo in China, Asus presented an exclusive ROG Astral RTX 5090D graphics card weighing over 7 kilograms, of which 5 kilograms consist of 24-karat gold. The gold is used on the cooler frame, backplate, and fins, where, in addition to the luxurious appearance, it also provides high thermal conductivity.
The material value of this card exceeds 500,000 dollars (approx. 460,000 €), while technologically it is based on the ROG Astral RTX 5090 model with 32 GB of GDDR7 RAM, PCIe 5.0, and a 2.61 GHz clock. Asus does not plan to sell the gold model; it is a PR and design experiment that may end up in a charity auction or be melted down after the exhibition. Public reactions are mixed: some admire the technical and artistic value, while others criticize the project as a symbol of excessive luxury at a time of rising hardware prices.
Source: https://videocardz.com/newz/asus-shows-how-they-built-the-worlds-most-expensive-gpu-a-5kg-gold-rtx-5090d-worth-half-a-million-dollars
September 2025
Apple is preparing major iPhone innovations
According to information from Mark Gurman, Apple is preparing a series of major changes. This year, the ultra-thin iPhone 17 Air is expected to arrive, designed to impress with extremely low thickness and weight (battery life may be sacrificed in favor of design). In fall 2026, the first foldable iPhone is expected to follow, with the codename V68, featuring a design similar to the Galaxy Z Fold, no physical SIM slot (eSIM only), a return of Touch ID, a new Apple C2 modem, in-cell technology to reduce crease visibility, and likely four cameras (front cameras on both the outer and inner displays plus two rear sensors).
A year later, in fall 2027, Apple plans a larger redesign of the entire lineup for the 20th anniversary of the iPhone: reportedly an entirely new appearance with rounded glass edges around the perimeter and a related redesign of iOS (Liquid Glass style). If these plans come true, Apple will significantly strengthen its design and product innovation efforts.
Source: https://techcrunch.com/2025/09/14/iphone-17-iphone-air-airpods-pro-3-and-everything-else-announced-at-apples-hardware-event/
Sam Altman warns that conversations with ChatGPT could be used in legal proceedings
OpenAI CEO Sam Altman warned in an interview that there is no clear legal framework protecting AI conversations from court requests or data leaks. He reminds users that questions and information shared with ChatGPT may in the future be subject to disclosure to authorities or legal processes, similar to how data from social networks is handled today. Therefore, deleting content by the user may not guarantee its complete removal from systems.
Altman also pointed out the risks of sharing sensitive information (medical records, therapy, legal advice), which, when communicated to an AI, do not have the same legal protection as when shared with professionals. He also expects that as AI usage grows, governments will increasingly seek access and oversight, making it necessary to address privacy and transparency rules. This results in practical advice for users to be cautious with sensitive or legally important information and not think of AI as a place that offers the same protection as professional confidential services.
Source: https://cointelegraph.com/news/dont-share-everything-with-chatgpt-sam-altman
Google Live Translate – live translation that also teaches you the language
Google is expanding Live Translate features in Translate so that, in addition to instant translation of spoken words, it also serves as an interactive language teacher. The feature, available in mobile apps for Android and iOS, plays the translation through the speaker, displays a transcript, and thanks to advanced AI voice models, takes into account conversational pauses, accent, and intonation while filtering background noise for use even in noisy environments.
New features include customized listening and speaking practice. The user enters their level and goal, presses the training button, and the system dynamically adjusts the difficulty of exercises based on usage. The feature is currently in beta and available to users in the USA, Mexico, and India. After fine-tuning, it is planned to be gradually released in additional regions.
Source: https://blog.google/products/translate/language-learning-live-translate/
PromptLock – the first ransomware powered by generative AI
ESET researchers have discovered PromptLock ransomware, which uses a generative language model to dynamically generate malicious scripts during attacks. The ransomware launches an LLM (locally or via a public API) and in real time creates Lua scripts compatible with Windows, Linux, and macOS. It is written in Go and uses the 128-bit SPECK encryption algorithm. PromptLock scans local files, analyzes their content, and based on predefined prompts decides which files to exfiltrate or encrypt. The destructive part of the code is already included in the program, though it has not yet been activated in all variants.
ESET considers PromptLock a proof of concept that significantly lowers the barrier for deploying sophisticated attacks. Instead of requiring a large team of developers, an attacker can use a well-configured AI model to automatically generate and adapt malicious code. Early variants have already appeared on VirusTotal. Researchers warn that such tools may make detection and defense more difficult. ESET has published technical details and classified the threat as Filecoder.PromptLock. They also recommend monitoring their channels for the latest indicators of compromise and mitigation steps.
Source: https://www.eset.com/sk/o-nas/press-centrum/eset-tlacove-spravy/eset-objavil-promptlock-prvy-ransomver-pohanany-umelou-inteligenciou/
YouTube tightens control over Premium sharing
YouTube has started actively enforcing family subscription rules and checking whether members of a family plan live at the same address. Some users have already received emails warning that if usage across multiple addresses is confirmed, their Premium subscription will be suspended or terminated. In such a case, they have 14 days to fix the issue. Previously, electronic address verification occurred once every 30 days, but the platform now monitors and penalizes violations more strictly.
This step mirrors the practice of other streaming services. Its goal is to reduce sharing between multiple households and increase the number of individual paying users. If this affects you, YouTube offers the option of resolving the issue through support; otherwise, you will need to switch to an individual subscription or ensure that all family members meet the same-address requirement. For the average user, this means checking family plan settings and considering legal sharing options or switching to a standalone subscription.
Source: https://www.androidauthority.com/youtube-premium-family-plan-restrictions-3593400/
Windows 11 25H2 will arrive in October 2025 – what’s new
Microsoft has made 25H2 available only to Windows Insiders in the Release Preview channel so far. Full release is planned for early October 2025. It is mainly an enablement package (eKB), a shared base of system files with a set of features that are only activated, so installation will be fast (in many cases just one restart) and compatibility with systems running version 24H2 will remain. Therefore, it will not offer users many visible UI changes or new features, but rather technical improvements and under-the-hood fixes.
Specific changes include the removal of outdated tools, mainly PowerShell 2.0 and the Windows Management Instrumentation command-line (WMIC), reflecting Microsoft’s effort to reduce legacy components and improve system security and maintainability. Support for 25H2 will be extended to 24 months for Home and Pro and up to 3 years for Enterprise, which may motivate both businesses and home users to update. However, features will be released gradually, meaning not everyone will receive them at the same time. Deployment will be managed by Microsoft, and some new features may be available only to selected user groups initially in Insider channels.
Source: https://blogs.windows.com/windows-insider/2025/08/29/releasing-windows-11-version-25h2-to-the-release-preview-channel/
Google will introduce mandatory verification of Android developers in Play
Google is launching a new Developer Verification system aimed at reducing the spread of malicious apps installed outside the Play Store. Identity verification will be tied to a unique DUNS identifier (a global business ID), and it reportedly already helps reduce malware presence. According to Google, apps installed outside its store are significantly more likely to contain malicious code. The system will not affect apps already in Play, and developers can gain early access as soon as October this year.
Deployment will take place gradually. Google plans mandatory verification for all developers starting in March next year, with regional rollout in some countries (Brazil, Indonesia, Singapore, Thailand) as early as September 2026 and global enforcement in 2027. The requirement applies only to certified devices with Google services (that passed CTS and have Play/Play Protect); devices without Google services (such as some Huawei models) will therefore be outside this regime. Google also clarified that the new policy will not universally exclude common non-custodial crypto wallets and that specific legal requirements (e.g., MiCA in the EU) for institutional apps will remain in effect.
Source: https://www.bleepingcomputer.com/news/security/google-to-verify-all-android-devs-to-protect-users-from-malware/
ChatGPT will launch special child accounts with parental controls
OpenAI is preparing special accounts for users aged 13 and older, which will work on the basis of invitations from parents and include expanded parental settings. Parents will be able, starting in October, to disable chat history and model memory, receive notifications when the conversation shifts toward sensitive topics (self-harm, suicidal thoughts, eating disorders, or drugs), and generally have greater insight into risky situations. OpenAI is working with doctors and psychologists on this, and within 120 days plans to introduce a parental feature. A separate parental section should appear in approximately one month.
Part of the measures also includes improving the handling of long conversations, which until now have worsened the ability of safety algorithms to conduct safe dialogue with children. The new child accounts are to be tied to safer versions of the model (for example GPT-5 with adjusted decision-making for more demanding topics), with OpenAI indicating that the system will be more sensitive in steering discussions, not in providing definitive answers to sensitive questions.
Several practical questions, however, remain open – whether and how child accounts will function given the possibility of free access to ChatGPT or temporary anonymous mode, and to what extent parents will be willing and able to supervise conversations. The feature also raises discussion about the balance between youth protection and the privacy or autonomy of young users.
Source: https://openai.com/index/building-more-helpful-chatgpt-experiences-for-everyone
Functionality of Internet browsers after the end of Windows 10 support
The end of mainstream support for Windows 10 will occur on October 14, 2025, but Microsoft has already announced extended support for some key components. Microsoft Edge and WebView2 Runtime will receive updates on Windows 10 22H2 at least until October 2028. This ensures that Edge itself and applications using WebView2 will remain secure and compatible for several more years, so regular browsing and the functionality of most web applications should remain unaffected.
Other browser vendors have not yet issued unified announcements. Opera has remained silent about its plans, Firefox is relying on its long-term support policy through ESR (for example, ESR 115 is supported until March of the following year) and Mozilla will also end support for 32-bit Linux systems in 2026 (the last version with this support will be Firefox 114, ESR until version 140 expected in September 2026). Vivaldi has not yet announced changes, with the only relevant announcement being the end of support for the old extension interface (Manifest v2), but that is not directly related to Windows 10. Overall, browsers will remain functional, but companies will gradually adjust support according to the system’s lifecycle and architectures (for example, 32-bit).
Source: https://www.ghacks.net/2025/09/04/mozilla-extends-firefox-esr-115-support-for-windows-7-until-march-2026/
ChatGPT introduces conversation branching for better idea exploration
ChatGPT is introducing a new feature that allows branching of conversations. A user can open a new branch at any point in the discussion and continue in a different direction, while the original line remains preserved. The feature can be activated in the menu under the AI’s response using the option “branch in a new chat”. The new branch will open in a separate tab and will contain the entire conversation history up to the moment where the user decided to branch off. This allows testing alternative ideas, scenarios, or prompts without risking disruption of the original discussion.
So far only a few tools offer a similar function, including ChatGPT and Venice.ai (where it is referred to as a fork). Other major AI chatbot platforms such as Gemini, Grok, or Perplexity do not yet provide branching. This new feature therefore gives users a more practical way to work with conversations and simplifies maintaining clarity when developing multiple lines of thought at once.
Source: https://lifehacker.com/tech/chatgpt-has-added-branching-chats
A new era of high-speed Internet begins in Slovakia
UPC has launched pilot connectivity with download speeds of up to 2500 Mbit/s and upload speeds of up to 150 Mbit/s in selected parts of Bratislava. The service is currently available only as a pilot, and the operator offers it as a demonstration of the technological maturity of its hybrid network. The packages now start at 600 Mbit/s and go up to 2500 Mbit/s. UPC is building on previous innovations, having introduced 1200 Mbit/s last year, and is now continuing to expand network capacity.
In practice, a 2.5 Gbit/s speed means instantaneous downloading of large files, smooth gaming, and stable simultaneous use of the network by multiple devices in the household. The 150 Mbit/s upload is sufficient for regular video uploads or cloud backups, although more demanding professional scenarios may require even higher uplinks. To fully use the speed, modern devices and infrastructure are recommended (quality CAT6/CAT7 cables, Wi-Fi 6 or newer, more powerful computers).
UPC states that the results were confirmed by independent measurements (Steam, nPerf) and that the first demo devices with Mini-SSD and gaming consoles demonstrated practical usability. The launch of the pilot phase is intended to be a step toward wider deployment in additional locations and part of an infrastructure investment aimed at bringing Slovakia closer to countries with the most advanced broadband networks.
Source: https://shop.upc.sk/sk/novinky/znamy-slovensky-operator-opat-prekvapuje-prinasa-raketovy-internet-az-2500-mbps
Drive manufacturers announce price increases, SSD and HDD prices will rise
SSD and HDD prices are expected to increase. The analytics company TrendForce has recorded pressure on the storage market and major players have already announced price hikes. SanDisk has signaled price increases of approximately 10%, and a similar move has been announced by Western Digital (SanDisk has been its division since 2016). The companies justify the changes by a combination of higher logistics costs, slower shipping times (delivery times are reportedly extending by 6–10 weeks), and rising demand.
Increasing demand for capacity is pushing prices upward primarily due to the growth of artificial intelligence projects and cloud services, and large cloud and data center providers are demanding increasingly larger drives (for example 32 TB), which also puts pressure on segments meant for desktops or small businesses. The result may be price increases not only for top-tier models, but indirectly also for more common capacities (2 TB and above), as storage size becomes the new standard.
Source: https://www.trendforce.com/news/2025/09/10/insights-memory-spot-price-update-sandisks-10-price-hike-meets-demand-side-resistance/
Microsoft extends security fixes for Office on Windows 10
Microsoft has adjusted its support roadmap for its office applications. Microsoft 365 will receive security updates on Windows 10 at least until October 10, 2028. Originally, the end of support was expected to coincide with the end of Windows 10 support on October 14, 2025. The extended support mainly covers fixes and security patches to keep users protected even on the older OS.
It is important, however, to distinguish between the Microsoft 365 subscription and one-time Office licenses: Office 2016 and 2019 will lose support in 2025, Office 2021 (LTSC) ends in October 2026, and the latest Office 2024 is supported until 2029. Be cautious when purchasing older licenses. Cheap offers on the market can be tempting, but older versions will no longer receive security updates, so a small extra cost for a newer license often means longer support and higher security.
Source: https://learn.microsoft.com/en-us/microsoft-365-apps/end-of-support/windows-10-support
Nvidia invests in Intel, GeForce RTX will become part of integrated CPUs
Nvidia and Intel have announced a strategic partnership under which Nvidia will purchase Intel shares worth 5 billion USD, which should enable deeper technical integration for both companies. The plan includes the development of new x86 chips for AI, the use of NVLink in data centers, and most significantly, integrated GeForce RTX graphics units will be used directly in Intel processors. The agreement would provide Intel with a financial injection and give Nvidia a faster path to a stronger position in x86 architecture and integrated solutions.
The consequences, however, are substantial and still unclear. Integrating RTX into iGPUs may strengthen gaming and acceleration performance of integrated chips, but at the same time threaten the future of Intel Arc and change the competitive position of AMD in both integrated and discrete graphics. Not all technical or regulatory details are known (the transaction requires approval from regulators). For example, it is not yet known to what extent this will be a complete replacement of Intel iGPUs, which segments will be served by NVLink, or whether Nvidia will begin manufacturing chips with Intel. After the announcement of the news, Intel’s stock jumped significantly, and Nvidia’s shares rose slightly. Negotiations reportedly took approximately a year.
Source: https://www.pcmag.com/news/shocker-intel-cpus-to-feature-built-in-nvidia-rtx-graphics
The first historical 1 Gbps laser data transmission across 5,500 km
General Atomics EMS together with Kepler Communications demonstrated a successful optical wireless connection between a De Havilland DHC-6 Twin Otter aircraft and a satellite in low Earth orbit at a distance of approximately 5,500 km. The aircraft was equipped with a 10 W Optical Communication Terminal (OCT) laser module in a 30 cm retractable LAC-12 turret. The system theoretically supports up to 2.5 Gbps, but in the test they achieved a practical data rate of around 1 Gbps while actively tracking the satellite at speeds of up to 25° per second.
This result represents one of the first reliable high-speed optical links between an aircraft and a satellite and has the potential to significantly improve data throughput capacity for commercial applications (passenger connectivity, HD/4K video transmission) as well as military or rescue deployments (real-time sensor transmission, ISR). Laser links offer the advantages of high data transfer speeds and low interference compared to RF, but practical deployment will depend on conditions such as precise aiming and stable tracking, sensitivity to atmospheric conditions (clouds, rain), and the need for integration with satellite networks or ground infrastructure.
However, open questions remain regarding operational reliability in poor weather conditions, safety and spectrum regulation, as well as the commercial economics of such solutions when deployed more widely.
Source: https://interestingengineering.com/military/plane-to-satellite-laser-beams
August 2025
The EU introduces the first comprehensive regulatory framework for artificial intelligence
The European Union has presented a legislative framework for regulating artificial intelligence, which is expected to take effect within two years. The legislation focuses on assessing systems according to risk levels, and unacceptable risks such as social scoring, cognitive manipulation, and biometric identification in public spaces have been banned since February.
The rules include special oversight of high-risk AI tools used in healthcare, transportation, infrastructure, and law enforcement. Another important element is mandatory labeling of content generated or modified by artificial intelligence, to increase transparency and user trust.
Source: europarl.europa.eu/topics/en/article/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence
Google launches AI Mode search
Google is launching a new AI mode in search globally, which fundamentally changes the way users obtain information. The feature was originally tested in the USA and India but is gradually expanding to additional countries. It is not yet available in Slovakia, though it has appeared in the beta version of the Google app for Android. AI Mode provides quick access to search through a new widget that can be easily activated on the smartphone home screen.
In addition, Google is testing AI insights that will allow users to explore their search queries more deeply through supplementary options. However, the new form of search also raises concerns because users may become passive consumers of information and rely solely on AI answers without critically evaluating sources. The integration of AI thus brings not only convenience and speed but also challenges in digital literacy and trustworthiness of information from the Internet.
Source: phonearena.com/news/google-just-made-its-ai-only-one-tap-away-on-android_id172670
AI Copilot in Microsoft Edge launched (also for Slovakia)
Microsoft has made the AI Copilot mode in the Edge browser available to users in Slovakia. It is activated simply by a toggle in the settings, turning Edge into an AI browser without requiring the installation of a new application.
The change is most noticeable in the new home screen with an input field like ChatGPT, which also functions as a traditional address bar. Copilot can, for example, summarize a YouTube video, though for now only based on its description. The features will gradually expand, and Microsoft emphasizes that this is a compromise between the regular Edge and an AI browser, which will be appreciated especially by users who actively use AI.
Source: microsoft.com/en-us/edge/ai-powered/copilot-mode?form=MG0AWI&cs=1155835057
Opera sues Microsoft over unfair practices in browsers
Opera has filed an official complaint against Microsoft with the Administrative Council for Economic Defense in Brazil, where it is the third most widely used browser. It claims that Microsoft uses manipulative design and “dark patterns” to force users to choose Edge, changes default settings (for example switching the search engine to Bing), and uses intrusive banners to discourage the installation of competing browsers. The issue also involves default opening of PDF files or links in Outlook.
Similar cases have already been addressed in Europe, where Microsoft was required under the DMA to allow the uninstallation of Edge. If Opera succeeds, Microsoft could be forced to allow the preinstallation of a browser selection tool, like the so-called ballot screen, although that did not have a significant impact on market share in the past.
Source: theverge.com/news/715082/opera-microsoft-competition-complaint-edge-windows-tricks
ChatGPT launches the “Study and Learn” mode
OpenAI has expanded ChatGPT with a new educational mode called “Study and Learn”, which is available even for free. The feature targets students and teachers and can explain school material, help with homework, or prepare tests and quizzes for practice.
Unlike the regular ChatGPT mode, this one focuses exclusively on learning. Instead of providing ready-made solutions, it is meant to serve as an interactive tutor. A student can choose the level of explanation, for example geography explained for high school students, and the chatbot will adapt. The goal is for the learner to understand or practice the material even when a teacher or classmate is not available.
The feature already works in Slovakia and should be available to all users regardless of subscription. If you do not see it in the menu yet, it will likely appear for you in the coming days.
Source: openai.com/index/chatgpt-study-mode/
The Play Store faces major changes following the dispute with Epic Games
Based on a court ruling in the long-running dispute with Epic Games, Google must adjust its Play Store within 14 days. The ruling confirmed that Google’s practices were anti-competitive, and therefore the company must open its ecosystem to competition. Key changes include allowing alternative payment systems in apps, giving developers the option to inform users about other payment methods, and enabling sharing of links to apps outside of the Play Store. Developers will also be able to set their own prices independently and Google may no longer pay manufacturers or carriers for exclusivity or for preinstalling its store.
A major step is also the requirement to cooperate with Epic Games and allow integration of competing stores directly into the Play Store environment. Users will therefore gain access to apps from multiple sources in one place. Google, however, warns that rapid implementation of these changes could destabilize the Android ecosystem and has therefore requested a delay in the ruling, although the chances of success are uncertain.
Source: theverge.com/news/717440/google-epic-open-play-store-emergency-stay
ChatGPT will change responses to sensitive questions, and usage duration warnings will be added
OpenAI has announced that ChatGPT will no longer provide definitive answers to personal or emotionally sensitive questions, such as whether a user should break up with their partner. Until now, the chatbot responded strictly and factually, which could come across as insensitive. From now on, it is intended to answer more neutrally, with greater consideration for users’ emotions and without directly influencing their decisions.
Another new feature is the introduction of a usage time control. If the system recognizes that communication has been ongoing for too long, it will alert the user and recommend taking a break. These changes aim to support more sensitive and healthier use of the chatbot, although OpenAI has not yet announced an exact release date.
Source: openai.com/index/how-we’re-optimizing-chatgpt/
Copilot 3D brings simple creation of 3D models from images
Microsoft has introduced the Copilot 3D feature, which makes it easy to create 3D models from ordinary images. All you need to do is upload a file in PNG, JPG, or SVG format, and the tool will automatically generate both a textured version and a mesh model that can be used for 3D printing, virtual reality, games, or animations.
Users can experiment even without a subscription, as long as they sign in to Copilot on the web. Models can be stored for 28 days, after which they are automatically deleted. Outputs are also available in GBL format, and it is possible to download or edit creations made by other users. The feature currently works only in 2D-to-3D conversion mode and has a 10 MB image size limit. Microsoft also points out that users are allowed to upload only images for which they own the copyright.
Source: copilot.microsoft.com/labs/experiments/copilot-3d
Microsoft Edge on Windows 10 will be supported until 2028
Microsoft has announced that the Edge browser and the WebView2 Runtime system component will receive support in Windows 10 22H2 at least until October 2028. This is a lifecycle extension similar to Windows 7, where browsers also continued receiving updates even after the official end of system support. For users, this means improved security and compatibility for applications that rely on WebView2 as a runtime module. It is a key element referenced by many third-party applications. Without its support, developers would have to redesign software for a different runtime module, which would bring unnecessary complications. Microsoft’s decision thus extends the lifespan of many applications that would otherwise lose full functionality.
The support applies to Edge starting from version 94, and for the average user this poses no issue, as the browser updates automatically and the current stable version is already 139. This means that most users will have support ensured without any additional steps.
Source: learn.microsoft.com/en-in/deployedge/microsoft-edge-support-lifecycle
Google Gemini brings personalized responses thanks to context memory
Google has introduced a new feature in Gemini that will allow the model to respond better based on previous interactions. The feature, called Temporary Chats, works by letting the system create short notes about the user to tailor responses to their preferences. Unlike permanent chat history, this is limited memory of details intended to improve future personalization. The feature will first be available in Gemini 2.5 Pro and later in the Flash version.
Personalization may show up, for example, in recipe recommendations, where Gemini remembers rejection of certain foods, or in programming, where responses can be tailored to a preferred programming language or environment. The feature is enabled by default, but users can disable it in the settings under Personal Context. A similar principle also works in other AI tools, such as Perplexity.
Google is also introducing the option of Temporary Chats, in which no user information is used. These conversations are separated from regular chats and are automatically deleted after 72 hours, offering the ability to converse without the influence of previous responses.
Source: blog.google/products/gemini/temporary-chats-privacy-controls/
China introduces a small SSD the size of a SIM card
The Chinese company Biwin has launched a new Mini SSD drive that measures only 15 × 17 × 1.4 mm and resembles the size of a SIM card. Despite its small dimensions, it offers high performance with read speeds of up to 3,700 MB/s and write speeds of 3,400 MB/s over PCIe 4.0 × 2, significantly surpassing current MicroSD cards. The drive is available in capacities of 512 GB, 1 TB, and 2 TB, and features IP68 certification for resistance to water, dust, and drops from heights of up to three meters.
The Mini SSD is designed with an integrated slot similar to a SIM card and was introduced alongside the GPD Win 5 portable gaming consoles and Super X hybrid devices at the ChinaJoy exhibition in Shanghai. If the technology catches on, it could significantly change memory expansion options in laptops, phones, tablets, or action cameras. However, there is currently no unified format standard for this type of drive, and its global availability and price are unknown.
Source: heverge.com/news/759624/china-is-about-to-launch-ssds-so-small-you-insert-them-like-a-sim-card
OpenAI considers adding ads to ChatGPT
OpenAI, the company behind the ChatGPT chatbot, is considering the possibility of including ads as a source of funding for further development of its models, including the latest GPT-5. However, there is currently no concrete plan for how the ads should work, since their presence could affect user trust and the quality of responses.
Potential forms of advertising include direct display between conversations or sponsored recommendations. For example, when asking about a restaurant, paid suggestions could appear at the top. ChatGPT is currently available for free and without ads, which significantly contributes to its popularity. An alternative monetization option is the Shop in ChatGPT project, which would allow OpenAI to receive a share of revenue from purchases made based on the chatbot’s recommendations. This concept is still under discussion and has not yet been implemented.
Source: goml.io/gen-ai-live/openai-may-add-ads-in-chatgpt
OpenAI defeats Grok in the finals of an AI chess tournament
In an unconventional chess tournament where general-purpose AI models designed for everyday use competed, OpenAI’s o3 model defeated Grok 4 from xAI, the company founded by Elon Musk. This tournament was not intended for specialized chess engines but for universal AI systems, testing these models’ capabilities in strategic thinking and decision-making.
The tournament took place on the Kaggle platform and included eight large language models from OpenAI, Google, xAI, Anthropic, and Chinese developers DeepSeek and Moonshot AI. Grok 4 appeared to be the favorite in the early rounds, but in the semifinals and finals it made several mistakes, including repeatedly losing its queen, which allowed o3 to claim a series of convincing victories. Google’s Gemini model took third place.
Chess and similar complex strategy games are often used as benchmarks for assessing AI models’ abilities in planning and reasoning. Such competitions help developers understand how well their systems can analyze rules, make decisions, and optimize outcomes in environments that may seem simple but require a high level of logic and strategy.
Source: chess.com/news/view/kaggle-game-arena-chess-2025-day-3
July 2025
Google Search receives an AI Mode
Google has officially launched its AI Mode in Search for Android and iOS in more than 100 countries, including India, marking the beginning of its global rollout. This mode uses the Gemini 2.5 model and a fan-out technique to break down a query into subtopics, process them in parallel, and display complex answers with links to sources.
AI Mode can handle queries up to three times longer than classic search, supports voice input and image queries through Google Lens, and allows users to ask follow-up questions directly in the interface. You can activate it with a single tap on the new icon in the Google search bar, and availability for Slovak users will be announced later.
Source: blog.google/intl/en-in/products/google-search-introducing-ai-mode-in-india/
Windows 11 will replace the Blue Screen of Death with a black one
Microsoft has decided to revive an almost 40-year-old icon of its operating system and will change the traditional Blue Screen of Death (BSOD) to a Black Screen of Death in the upcoming update. The new screen will continue to display detailed error information, including the code and the name of the problematic driver, but in a dark visualization that also matches the existing black screens shown during system updates or startup.
According to David Weston, Microsoft’s Vice President for Security, the change is meant to help users better focus on the root of the problem and more clearly identify the cause of the failure. The adjustment also responds to last year’s CrowdStrike update incident, when massive BSODs brought entire organizations to a standstill, including hospitals and airports, and it highlights Microsoft’s efforts to improve readability and place more emphasis on error diagnostics.
Source: theverge.com/news/692648/microsoft-bsod-black-screen-of-death-color-change-official
Microsoft claims that Windows 11 is more powerful compared to Windows 10
Microsoft states on its website that Windows 11 offers significantly faster task processing and longer battery life compared to Windows 10, for example up to 2.7 hours more during video playback and up to 2.3 times faster benchmark results. However, these tests are based on comparing newer hardware (Intel 12th and 13th generation) with previous generations (6th to 10th), which indicates that a hardware upgrade is important to achieve the full performance of Windows 11.
Microsoft is offering a free one-year extension of Windows 10 support, after which it plans to introduce paid extended support and encourages users to switch to Windows 11 along with newer hardware. Critics warn, however, that older devices without official support will remain vulnerable and that a focus on new features should go hand in hand with providing fixes for existing users.
Source: learn.microsoft.com/en-us/windows/deployment/performance-lab/claims-disclosures?tabs=windows-11
JPEG and PNG receive modern features – animations, HDR, and the new JPEG XL
The PNG and JPEG formats, which have dominated the Internet for nearly a quarter of a century, have received major upgrades. PNG is gaining animations and extended color-gamut support for HDR images, with the new specification (v0.90) already working in major browsers and creative applications. Work is also underway on a fourth generation with better HDR/SDR interoperability and a fifth generation with updated compression.
JPEG, on the other hand, is arriving in a new form as JPEG XL (JXL), offering both lossy and lossless compression with up to 60% file-size savings over classic JPEG at the same quality. JXL supports reverse conversion, allowing images to be losslessly converted back from JXL to the original JPEG, and it adds animations, making it competitive with formats like GIF or APNG. Although support for JPEG XL is still experimental (Chrome, Firefox), it is a promising way to keep older formats relevant even in the era of WebP and AVIF.
Source: w3.org/TR/png-3/#12Encoder-colour-handling
Google Photos will allow the creation of short videos
Google Photos is preparing a feature called Remix that uses generative AI to transform static images into short animated videos. After selecting a photo, you choose the animation style in the testing interface, ranging from subtle movements to dramatic effects, and the application renders a smooth sequence for you. The result is generated within one minute and allows memories to be brought to life in a new, dynamic way.
Remix is currently in internal testing: the interface offers large buttons for selecting the effect and shows a preview at first launch, but some users still encounter error messages. If Google approves the release, the feature will become an attractive tool for those who want to further enhance and share their photos in a modern animated form.
Source: androidauthority.com/google-photos-remix-interface-update-3575599/
Windows 11 is the most used desktop system
In July 2025, the Windows 11 operating system surpassed Windows 10 and reached a 51.77% share of the desktop operating-system market, while Windows 10 dropped to 45%. Nearly four years have passed since the official completion of Windows 11 (June 24, 2021), although it did not become widely available until October 2021.
Migration in large organizations is accelerating mainly due to the end of free support for Windows 10, while households are adopting it more slowly. Among gamers, Windows 11 has been the number one system since last September: in June 2025 it holds a 59.8% share, Windows 10 has 35.7%, Linux 2.6%, and macOS 1.8%.
Source: techpowerup.com/338670/windows-11-finally-overtakes-windows-10-as-the-most-dominant-operating-system
Grok 4 – Musk’s Big Bang in AI
The company xAI, backed by Elon Musk, has introduced the Grok 4 model, which claims to outperform GPT-5 and Claude 4 Opus in both performance and logic. Musk compared its launch to a “Big Bang in AI” and declared that Grok 4 is smarter than all postgraduate students in every field at once, allowing it to discover new technologies even before researchers become aware of them.
Grok 4 was trained on the Colossus supercomputer and brings stronger reasoning, multimodal inputs (text and images), and a specialized Grok 4 Code variant for writing and debugging code. The model retains access to up-to-date information through DeepSearch from the X platform and is tuned to understand internet memes and humor. The company xAI skipped version 3.5 to emphasize the major leap forward compared to the competition.
Source: x.ai/news/grok-4
LPDDR6 memory will bring double the speed and lower power consumption
The JEDEC organization has approved the LPDDR6 standard, which brings effective frequencies from 10,667 MHz up to 14,400 MHz (compared to 8,533 MHz for LPDDR5X), corresponding to a bandwidth of 28.5 to 38.4 GB/s per chip. The new 24-bit architecture (2 × 12-bit subchannels) replaces the previous 32-bit frames (2 × 16-bit), reduces latency, and saves energy thanks to dynamic frequency scaling and reduced voltage.
LPDDR6 is expected to appear in the first smartphones, tablets, and ultramobile laptops as early as 2026. Major memory manufacturers Samsung, Micron, and SK Hynix are already finalizing preparations for mass production in order to take advantage of the twofold performance increase compared to LPDDR5 while lowering power requirements.
Source: jedec.org/news/pressreleases/jedec%C2%AE-releases-new-lpddr6-standard-enhance-mobile-and-ai-memory-performance
AI consumption compared to a microwave
A study from MIT warns that generating a 5-second video using AI consumes approximately 3.4 MJ, which corresponds to one hour of operating a microwave oven. An average response from a large language model such as Llama 3.1 405B requires around 3,353 J (about 8 seconds of microwave use).
However, exact numbers are often non-transparent and vary by provider. It is estimated that by 2028, up to half of the energy consumption in data centers could be attributed to tasks related to AI. Growing automation of routine processes will further increase these demands, especially if data centers continue to run primarily on fossil fuels.
Source: mashable.com/article/energy-ai-worse-than-we-thought
The most visited website in the world last year is Google
According to Visual Capitalist data from November 2024, Google reaches an average of 136 billion monthly visits, maintaining its position as the most visited website in the world. In second place is YouTube, where users spend an average of 25 minutes per visit. Facebook ranked third.
Also entering the top ten are Reddit (91 million monthly visits), ChatGPT (400 million weekly, 9th place), and Wikipedia, which serves as a key source of educational material for AI. The dominance of Google and Facebook covers roughly 90% of the advertising market, but the rise of AI tools and the popularity of platforms like Reddit suggest potential changes in future rankings.
Source: visualcapitalist.com/most-visited-websites-in-the-world/
A human (likely for the last time) defeated AI in a ten-hour programming marathon
At the prestigious AtCoder World Tour finals in Tokyo, top human programmers faced off against OpenAI’s OpenAIAHC model for the first time, and the winner was 42-year-old Polish programmer Przemysław “Psyho” Debiak, who achieved a result 9.5% better than the AI. While the model excelled at raw computation, Debiak used creativity, heuristics, adaptability, and intuition, which pure computational power cannot replace.
Debiak, a former OpenAI developer, worked exclusively in Visual Studio Code and admitted that competing with AI pushed him to deliver his absolute best, even though he had slept only ten hours in the three days leading up to the competition. Despite this triumph, AI is advancing rapidly in programming tasks — its success rate rose from 4.4% in 2023 to more than 71% in 2024. Tools like GitHub Copilot are used by up to 90% of developers today.
Source: tomshardware.com/tech-industry/artificial-intelligence/polish-programmer-beats-openais-custom-ai-in-10-hour-marathon-wins-world-coding-championship-possibly-the-last-human-winner
June 2025
Windows 11 news awaits
Microsoft is testing new options in the explorer’s contextual settings, such as linking to the Git cloud repository, displaying virtual environments (Hyper-V, Windows Sandbox), and expanded basic file management settings. In search, there will be information about the indexing status from different sites and whether a folder is available locally or only online (for example, from OneDrive). The explorer in the developer version also includes an AI actions submenu that allows basic image operations via the Sketchpad and Photos apps.
In addition, Windows 11 modifies the accessibility bar by displaying more settings in the floating bar and the quality of image compression within Windows Share (3 preset levels). It also introduces improved power saving mode logic, which activates when idle and quickly transitions back to a more powerful state when the user is active again. All of these new features are in the developer channel for now and will allow IT professionals and advanced users to customize how they work with the system in more detail.
Source: blogs.windows.com/windows-insider/2025/05/19/announcing-windows-11-insider-preview-build-26200-5603-dev-channel/
Xiaomi brings a smartphone that smells like coffee
Xiaomi has introduced the CIVI 5 Pro in a special brown finish called Iced Americano, which smells subtly of coffee, giving an exclusive impression without a strong perfume intensity. The device features a Snapdragon 8s Gen 4 chipset, a 6.55-inch OLED display with a 120Hz refresh rate, a slim body with a thickness of 7.45mm and a weight of around 181 grams in the coffee version.
Apart from Xiaomi, several manufacturers such as Motorola and Infinix are trying the trend of scented smartphones. Xiaomi CIVI 5 Pro is available only in China for now. The question remains whether the scented version will also reach Europe.
Source: androidauthority.com/iced-americano-phone-smells-like-coffee-3560711/
Edge as a gateway to AI models for developers
Microsoft is experimenting with allowing developers to access language models, specifically the Phi-4-mini model (3.8 billion parameters) via the API through the Edge browser. A prerequisite for using this functionality is the presence of Microsoft Edge on the device, with the goal of a cross-platform solution including macOS.
The feature is still experimental and available in the Edge Canary and Dev channels. It will allow easy integration of AI capabilities, such as text generation or summarization, directly in web applications without the need for separate libraries. Chrome offers a similar approach. Both companies are thus competing for developers’ attention to have their browser serve as the gateway to users’ models and data.
Source: theverge.com/news/669528/microsoft-ai-edge-browser-web-app-build-apis
Opera is preparing a new browser with AI
Opera is preparing a new dedicated browser with AI that will work as an agent. It will also offer users advanced options for editing code, creating apps or processing images based on text commands. Meanwhile, a concept presentation and an open waiting list are underway, with the goal of complementing the classic browser for both casual and expert users.
The service is expected to be subscription-based, but details on pricing and models are not yet known. Opera has experimented with AI in the standard browser in the past, including with the unfinished Opera Neon project from 2017, but this time it’s a standalone product focused on AI features integrated directly into the browser.
Source: operaneon.com
Google Photos celebrates 10 years with new AI editor
Google Photos is celebrating its 10-year anniversary, and to mark the occasion, Google has unveiled a revamped editor. This unifies the tools in one place and builds heavily on AI editing suggestions for the whole photo or specific parts of it. Previously exclusive to Pixel devices, features like Reimagine (enhance or change the scene) and Auto Frame (suggest an ideal crop and fill in missing parts) will soon be made available to all smartphone users.
Also new is the AI Enhance feature, which automatically generates three different editing suggestions (such as sharpening or removing objects) and the user chooses their preferred one. Another upcoming new feature is album sharing via QR code, which will allow instant sharing or adding photos between people in the vicinity. Google will start rolling out the new editor in June for Android (starting with version 8) and an iOS version will be available later in the year.
Source: androidauthority.com/10-years-google-photos-never-looking-back-3561775/
Play Store to offer AI assistant answering questions about apps
Google has started to gradually integrate the ability to ask questions to the Gemini AI directly in the app’s profile into the Play Store. Users will see a section for interacting with the AI, where they can get information about the app’s usage, settings, or rules without having to look elsewhere for answers.
The feature is available for now in the experimental version of the Play Store 46.1.39-31 and is launched gradually for selected users. Gemini will also suggest FAQs, making it easier to navigate and save time. As the answers are generated by the AI, they may not always be accurate, so it is recommended to verify them. The AI-supported section will only appear in apps that Gemini has sufficient information about, and will provide users with an assistant to help them navigate and understand the functionality of the apps.
Source: androidauthority.com/google-play-store-ask-play-about-this-app-rolling-out-3562259/
AI startup Builder.ai used 700 Indian engineers to answer
London-based firm Builder.ai, which presented itself as an AI-powered app development platform, has collapsed after it was discovered that it was actually using around 700 Indian engineers to manually perform tasks attributed to AI.
The startup also engaged in round-tripping with VerSe Innovation, billing for fictitious services to artificially boost revenue, which inflated actual revenue from $45 million to $180 million for 2023. After one of its major lenders, Viola Credit, foreclosed on $37 million due to breaches of loan terms and the company failing to pay employees, the CEO declared bankruptcy. The case highlights the risks of investing in startups that overestimate their technology capabilities and fail to manage transparency.
Source: business-standard.com/companies/news/builderai-faked-ai-700-indian-engineers-files-bankruptcy-microsoft-125060401006_1.html
IBM plans to develop Starling quantum computer with remarkable performance
IBM has announced an ambitious plan to develop the Starling quantum computer by 2029, which will have 200 logical qubits and the ability to perform up to 100 million quantum operations. The company says that simulating such a machine on classical supercomputers would require a quintillion powerful devices, illustrating the enormous potential of quantum computing.
The development is gradual. Already this year, the Nighthawk chip and the Quantum Loon platform are expected to be unveiled to show the first real quantum advantage. A modular Kookaburra chip for scalability is due to arrive in 2026, and a Cockatoo chip that interconnects multiple quantum units in 2027. Starling is set to tackle challenging tasks in drug development, new materials or molecular structure simulations in 2029. Beyond 2033, a Blue Jay platform with up to 2,000 logic qubits and the ability to perform more than a billion operations is planned. This incremental approach shows how quantum chips are gradually moving closer to real-world deployment outside the lab.
Source: tomshardware.com/tech-industry/quantum-computing/ibm-is-building-a-large-scale-quantum-computer-that-would-require-the-memory-of-more-than-a-quindecillion-of-the-worlds-most-powerful-supercomputers-to-simulate
AI tools start showing ads and sponsored recommendations
Recently, there have been instances of ChatGPT and other AI services inserting recommendations or sponsored links into replies, as confirmed by a user post on Reddit. A similar approach has already been implemented by tools such as Perplexity, where AI offers products or services as part of search results. OpenAI has hinted at the possibility of future ad revenue, particularly due to the ratio of paying to non-paying users, and like other platforms is experimenting with monetising AI through ads or premium subscription models.
This change reflects the desire of companies to maintain or increase revenue with a growing group of users who still use basic versions of free services. Advertising elements can take the form of directly embedded chat recommendations or pop-ups showing relevant products. Therefore, even when using AI chats, users need to be prepared that responses may include commercial propositions and verify information independently, similar to conventional advertising models on the internet.
Source: adguard.com/en/blog/chatgpt-ads-introduce-openai-blocking.html
ToS;DR translates the terms and conditions into an understandable form
The Terms of Service; Didn’t Read (ToS;DR) project collects and analyses the terms and conditions of the most popular Internet services to highlight problematic passages and explain them in “human language”. Volunteers rate each clause according to its approach to users’ rights, marking them with colour-coded categories from A (best) to E (unsafe), and highlighting situations where services are abusing data or storing content you think you’ve already deleted.
The service has been available for more than a decade and helps people quickly identify which conditions are worth a cautious reading or consideration of alternatives. Among the platforms with the weakest ratings are large platforms such as Facebook and YouTube, which often gain extensive access to user data in terms and conditions. Privacy-focused services such as Startpage or Tor Browser get better ratings. ToS;DR thus makes it easier to navigate behind the scenes of digital services and helps users make more informed decisions.
Source: tosdr.org/en
Google Maps will support greener summer travel
Google Maps is expanding features that use AI to recommend walking or public transport when speeds are comparable to a car. They will extend the prediction of such alternatives to new European cities, avoiding tens of millions of car journeys.
For cyclists, detailed data on cycle routes in other cities, information on dedicated lanes, traffic and gradients will be added. There will also be routes with low fuel or energy consumption and low emission zone alerts available in more than a thousand locations in Europe.
Source: blog.google/products/maps/new-maps-updates-for-sustainable-travel-europe/
Cloudflare fended off record DDoS attack
Cloudflare successfully repelled the largest DDoS attack on record, in which an attacker flooded a target IP address with data at 7.3 Tbps and sent 37.4 TB in 45 seconds, the equivalent of approximately 9,350 HD movies. The attack used massive UDP floods targeting all ports and reflection (amplification) techniques where spoofed requests forced legitimate servers to send responses directly to the victims, multiplying the load.
In recent years, similar extreme attacks have been on the rise: peaks of around 3.47 Tbps in 2022, 5.6 Tbps last year, and 6.5 Tbps lasting 49 seconds in April this year. Even when advanced systems are used to detect and deter threats, such massive reflex attacks form botnets of compromised devices and remain a relatively cheap means of extorting or testing infrastructure resilience. Continuous development of defense mechanisms is therefore crucial to maintain the availability of online services.
Source: tomshardware.com/tech-industry/cyber-security/massive-ddos-attack-delivered-37-4tb-in-45-seconds-equivalent-to-10-000-hd-movies-to-one-victim-ip-address-cloudflare-blocks-largest-cyber-assault-ever-recorded
Microsoft removes outdated drivers from Windows Update
Microsoft is starting to gradually purge old drivers from Windows Update that have replacements and may be a vulnerability or risk. In the first phase, drivers that have expired and are not tied to any current system update will be removed.
The process will be phased in: after support expires, there is a 6-month window during which partners can roll the driver back into Windows Update if necessary, with appropriate justification and Microsoft approval. If there are no objections, the drivers will be permanently removed. The goal is to ensure that the update service offers only the optimal and supported driver sets for current devices.
Source: techcommunity.microsoft.com/blog/hardwaredevcenter/removal-of-unwanted-drivers-from-windows-update/4425647
May 2025
ChatGPT Deep Research is now available without a subscription
OpenAI makes the advanced Deep Research feature available to a wider range of ChatGPT users, including those without a paid subscription. This feature allows the chatbot to generate complex analyses and thoughtful answers to more challenging questions.
Non-paying users have 5 tasks per month in the lightweight version of the feature. The largest number of tasks is available to subscribers of the Pro version (125 tasks per month plus 125 in the light version), followed by the Plus and Team versions (10 + 15 tasks). ChatGPT Edu users can use Deep Research 10 times a month, but without the light version. The feature is now also available for Slovak users. Look for the Deep Research menu in the interface.
Source: androidauthority.com/chatgpt-deep-research-lightweight-expansion-3548045/
ChatGPT can also help you with shopping
OpenAI has introduced a new feature within the GPT-4o that allows ChatGPT to act as a shopping assistant. It will provide users with personalized product recommendations including images, prices, reviews and links to e-shops in categories such as fashion, beauty, home and electronics.
The feature is available to all users for free, and even without logging in. OpenAI assures that the designs will not contain advertisements and the company will not profit from the feature. ChatGPT thus extends the use of text generation and Deep Research to practical help with online shopping. OpenAI is making the feature available gradually and it should be ready to try out in your interface soon.
Source: gizmochina.com/2025/04/29/chatgpt-shopping-features-launched/
Google has deleted almost half of the apps from the Play Store, the reason is the tightening of quality
Google has reduced the number of apps in the Play Store from more than 3.4 million to around 1.8 million since the start of 2024. The company has introduced stricter quality rules that ban broken, spammy or misleading apps and strengthened developer verification, new account testing and manual checks.
In 2024, Google not only removed millions of already published apps, but also blocked 2.36 million new titles before they were published. It also cancelled more than 158,000 developer accounts that were linked to malicious content. Despite this purge, more than 10,000 new quality apps were added this year. The goal is to provide users with safe and valuable content free of spam and useless apps.
Source: androidauthority.com/google-play-store-apps-count-may-2025-3549299/
Microsoft is getting more expensive – higher prices for Xbox, games and controllers
From 1 May 2025, Microsoft has increased the prices of its game consoles and accessories worldwide, including Slovakia. The base version of the Xbox Series S now costs from €350, while the Xbox Series X starts at €550 and the 2TB model up to €700. Prices for the official controllers range from €65 for the standard model to €150 for the Xbox Elite Wireless Controller Series 2.
Along with the hardware, Microsoft plans to raise the price of selected new games from the usual $70 to $80 by the end of the year, an increase of around €10 to €15 in Europe. The company justifies the move on the grounds of rising development costs. Monthly fees for Xbox Game Pass remain unchanged.
Source: support.xbox.com/en-US/help/hardware-network/console/may-2025-pricing-updates
Microsoft introduces password-free sign-in for new users
Microsoft now deactivates passwords on new accounts by default and only offers passwordless authentication methods such as Authenticator, biometrics and passkeys. For existing users, it allows you to remove your password in your account settings and switch to modern forms of sign-in.
The move follows a long-standing effort to eliminate the weaknesses of traditional passwords and increase account security. Passwordless login reduces the risk of attacks through password theft, phishing or password reuse and is becoming standard for new Microsoft accounts.
Source: microsoft.com/en-us/security/blog/2025/05/01/pushing-passkeys-forward-microsofts-latest-updates-for-simpler-safer-sign-ins/
Skype has ended, what happens to your account?
On May 5, 2025, Microsoft discontinued Skype. It will automatically move your contacts, call history and chats to the free Microsoft Teams version, without you having to create a new account.
The web.skype.com domain will be redirected to Teams, where you log in with your Skype account and access your existing data. If you don’t want to move to Teams, you can export your Skype data, including archived conversations, until January 2026. After that date, all your Skype data will be permanently deleted. Skype for Business is not affected by the change.
Source: support.microsoft.com/en-sk/skype/skype-odch%C3%A1dza-do-d%C3%B4chodku-v-m%C3%A1ji-2025-%C4%8Do-you-need-know%C5%A5-2a7d2501-427f-485e-8be0-2068a9f90472
Chrome for Android warns of dangerous notifications
Google is introducing a new malicious notification detection system to the mobile version of Chrome that uses artificial intelligence to scan incoming notifications. If Chrome identifies a potentially spammy or fraudulent notification, it will automatically offer you the option to unsubscribe from that page without having to manually search through your settings.
The feature responds to the increasing misuse of web notifications for fraud and unsolicited advertising. Google is using it to strengthen security for Chrome users on Android devices and is preparing it as part of a broader suite of protection tools, including an enhanced security mode for Android.
Source: androidauthority.com/chrome-android-web-notification-warning-3555382/
Google switches to new logo after ten years
Google is changing its iconic “G” logo, which last underwent a major redesign in 2015. The new version retains the original shape and colour palette, but instead of a strict transition between red, yellow, green and blue, it uses a gradual transition between colours that blend together. The logo is inspired by Gemini’s AI tool design style and reflects the company’s transition to AI capabilities.
The changes have so far appeared in the beta version of the Google Search mobile app on Android and iOS, and will gradually roll out to stable releases of other Google apps such as Maps and Chrome. Although this is more of a subtle cosmetic tweak from a user perspective, Google is signaling a new graphical direction and strengthening its position in AI with this change.
Source: 9to5google.com/2025/05/12/google-icon-update/
Teams will disable screenshots in conversations
Microsoft plans to add a “Prevent Screen Capture” mode to the Teams app starting in July 2025 that will block unauthorized screen capture in group calls. Participants will see a black screen when they try to take a snapshot, protecting sensitive information in the video conference.
The feature will be optional and available on Windows, Mac, Android, iOS and in the web interface. Unsupported platforms will only be able to work in audio meeting mode. This news fits into a broader trend of content protection in apps – like WhatsApp, it restricts media export and makes it harder to take unauthorized screenshots, though it won’t completely prevent taking photos using external devices.
Source: microsoft.com/en-us/microsoft-365/roadmap?id=490561
Copilot will also be voice activated
Microsoft has added to the Windows Insider version of Copilot the ability to activate the assistant by voice with the command “Hey Copilot”. After installing version 1.25051.10.0 of the application, you can activate this feature via the Alt + Spacebar keyboard shortcut. After saying the activation phrase, you will see a floating interface with a microphone that informs you that you are listening. If it doesn’t pick up any sound for a few seconds, it will automatically switch off.
Microsoft emphasizes that the entire pre-processed voice input 10-second buffer is processed on the device and not sent to the cloud. Only then is the audio data sent to the cloud, where in-depth analysis of the commands takes place. This reduces the amount of sensitive information transmitted and ensures user privacy.
Source: blogs.windows.com/windows-insider/2025/05/14/copilot-on-windows-hey-copilot-begins-rolling-out-to-windows-insiders/
YouTube ads are about to get even more annoying
YouTube plans to deploy a new way of serving ads, where Google’s artificial intelligence, Google Gemini , will search for the most attractive or engaging moment in a video and insert an ad spot there. The intention is to draw more attention to ads, although this may further disrupt viewing, especially for users without subscriptions.
It is not yet clear whether these ads will be skippable, nor exactly how they will be presented. However, the new ad placement is expected to increase users’ motivation to switch to YouTube Premium, where ads are not shown at all. The price of Premium ranges from €4.99 for a student plan to €15.99 for a family plan.
Source: androidauthority.com/youtube-peak-points-ads-3557905/
Google brings a new AI mode to search
Google has introduced an AI Mode shortcut to its Android and iOS mobile apps, replacing the existing icons for voice search and Google Lens. The new button at the top of the search bar carries the Gemini Assistant star symbol and gradient colors from the freshly unveiled logo.
The feature will allow the AI assistant to be launched directly and quickly from search without the need to switch to other modes. With the launch of AI Mode, Google has also moved Voice Search and Lens in the interface a level lower to simplify the top bar. AI mode will give users access to advanced generative intelligence capabilities right in their normal search for information. The new feature is currently rolling out gradually to stable versions of the app, while Google is also beta testing a widget-based invocation of AI mode on the home screen.
Source: 9to5google.com/2025/05/16/google-ai-mode-search-bar-android-ios/
Windows Explorer gets AI actions for quick image editing
In the developer version of Windows 11, a new AI Actions submenu has been added to Explorer, offering four tools for working with images directly from the context menu. The first option triggers a visual search via Bing, the second one-click blurs the background, the third removes unwanted objects, and the fourth cuts the object out of the background.
These actions make use of the built-in Paint and Photos apps, so that once selected, the image is opened in the respective app for final tweaking. Supported formats are .jpg, .jpeg and .png, allowing for quick basic editing without the need to run external graphics programs.
Source: blogs.windows.com/windows-insider/2025/05/19/announcing-windows-11-insider-preview-build-26200-5603-dev-channel/
Google AI will make more features available for free
At the I/O 2025 conference , Google announced that many AI features will be available without a paid subscription. The Gemini Live app will be expanded in the Android and iOS versions to include free tools for on-screen or real-time content recognition using the camera without the need for a subscription.
At the same time, Google introduced new paid packages within Google One – Google AI Pro (€21.99 per month) and Google AI Ultra ($249.99 per month, initially in the US). The Pro plan includes 2TB of storage, access to Gemini Pro and NotebookLM Pro, plus AI linking to personal documents or emails. Ultra adds video generation via Veo 3 and preferred use of Deep Think 2.5 Pro.
Other new features include the Imagen 4 model for image generation and an interactive Canvas mode that allows code editing, infographics, quizzes or audio summaries to be created directly in the AI tool, making the work of both programmers and content creators easier.
Source: blog.google/products/gemini/gemini-app-updates-io-2025/
Google Veo 2 expands AI video capabilities
Veo 2, Google’s AI-powered short video generation tool, has gained new features for more detailed scene inspection. Users can define characters and objects, choose a consistent style even when combining cartoon elements with photos, and adjust the aspect ratio of the video, for example, to extend a portrait to a reclining position.
In addition, camera controls have been added with motion and angle settings, as well as the ability to add or remove objects from an existing composition. These extensions are available to all AI package subscribers (Ultra in the US) and the limitations on the number of videos generated per month remain the same.
Source: blog.google/technology/ai/generative-media-models-io-2025
April 2025
Sony will increase the quality of photos – preparing 200 Mpx sensor
Sony is preparing a new 200 Mpx image sensor, which will have a higher resolution than previous solutions and will bring better low-light performance, more detailed images and better zoom. The sensor will not be exclusively used for its own models, but will also be available for other smartphone manufacturers. By focusing on higher image quality, Sony wants to compete with Samsung, which dominates in the high-resolution field. With a 1/1.3-inch sensor size and advanced technologies, the news is expected to improve autofocus, night vision and noise reduction. The first flagship models with this sensor could appear at the end of 2025.
Source: notebookcheck.net/Sony-s-new-200-MP-sensor-to-crush-the-opposition-and-be-adopted-by-next-gen-flagships.985233.0.html
Gmail improves email search with AI
Gmail is now using AI to improve search results. It offers users the ability to toggle between seeing the most relevant and most recent emails, increasing the chances that the email they want will show up right at the top. The new feature is available in the web version as well as in the Android and iOS apps. When assessing relevance, the AI takes into account, for example, topicality, frequently clicked messages or frequent contacts. The feature is gradually being rolled out to all users and will later be available for corporate accounts as well.
Source: blog.google/products/gmail/gmail-search-update-relevant-emails/
Gemini AI will make programming easier with new Canvas feature
Google introduced the Canvas feature within Gemini, which allows for easier editing of code and text. Canvas provides a clear interface with a question and answer split, and the resulting code can be easily copied and edited. The feature is available for Gemini and Gemini Advanced and works with the 2.0 Flash model. It allows you to mark specific sections of text to adjust tone, length or format. Programmers will appreciate the ability to test HTML and React codes in the interactive preview. Created documents can be exported to Google Docs.
Source: blog.google/products/gemini/gemini-collaboration-features/
ChatGPT improved image generation in the GPT-4o model
OpenAI has improved the ability to generate images directly in ChatGPT, including support for editing images directly in the chat interface. The feature uses the GPT-4o model, which better understands assignments and can work with multiple objects at once. Users can generate images without logging in, but get more options after logging in. The images contain C2PA information that records the origin and edit history. While the feature is not yet perfect, especially for complicated posters or text, it greatly expands the graphic creation capabilities in ChatGPT.
Source: openai.com/index/introducing-4o-image-generation/
Windows 11 and Copilot – Microsoft unveils roadmap of new features
Microsoft has published a clear roadmap of upcoming features for Windows 11 and Copilot Assistant to provide a transparent view of the system’s evolution and prepare IT professionals for the upcoming changes. The planned new features include features such as automatic tab refresh in Explorer, web widgets for developers, and improved CPU load calculation in Task Manager. Each feature in the table is labeled according to completion and availability status – for example, for the Insider program only or in the broader rollout phase. Among the features already available, for example, is Cocreator in the Copilot+ Device Sketchpad.
Source: microsoft.com/en-us/windows/business/roadmap
Gorilla Glass Ceramic – the next generation of display protection
Corning has introduced a new display protection technology – Gorilla Glass Ceramic, which uses ceramic materials to increase durability and has the ambition to replace existing variants such as Victus or Victus 2. The new Gorilla Glass Ceramic has an increased proportion of ceramic and can withstand repeated drops from a height of one metre onto a hard asphalt-like surface. This is in response to the frequent damage to displays, which is one of the most common reasons for replacing phones. The first device with this technology is expected to come from Motorola, and ceramic solutions are expected to become standard in the most premium models.
Source: androidauthority.com/corning-gorilla-glass-ceramic-3539449/
Bill Gates on AI: Three professions that won’t lose their jobs (yet)
Bill Gates predicts that artificial intelligence will dramatically change the job market in the next decade, but some professions remain safe for now. These are programmers, energy engineers and biologists.
Gates points out that while AI will replace many tasks, it still needs humans to be creative, solve complex problems and oversee its development. Programmers are important for creating and debugging code, power engineers work with overly complex systems, and biology still requires human inventiveness for research and discovery.
Source: windowscentral.com/software-apps/bill-gates-says-ai-will-replace-humans-for-most-things
Microsoft Office will speed up start – first in Word
Starting in May, Word will start up faster thanks to the Task Scheduler initializing the application at startup. A new feature called Office Startup Boot will allow Word to load faster by starting its initialization components as soon as the computer is turned on. It will only activate on devices with at least 8GB of RAM and 5GB of free disk space. The feature is optional and will not run when power saving mode is enabled. Although it would be more suited for Excel, Microsoft is testing it in Word first.
Source: zdnet.com/article/your-microsoft-office-apps-will-soon-launch-faster-but-the-speed-boost-is-optional/
Google Maps steps up fight against fake reviews
Google is improving its systems to detect fake reviews using artificial intelligence and Gemini technology, with measures being rolled out globally. Google Maps has long sought to curb review abuse (such as buying five-star reviews without visiting the establishment) and editing business profiles. In 2024, the company has already removed more than 240 million fake reviews and 12 million fake profiles, most of which were detected by algorithms before they were published. The new Gemini-based AI model can identify fraudulent behavior based on several new patterns of abuse, helping to block thousands of suspicious edits to company profiles, and can increase the credibility of reviews.
The new security measures are already in place in the US, UK and India, but will be rolled out globally from next month. Users will be alerted to removed reviews and suspicious profiles. With these measures, Google wants to ensure that Maps remains a trusted tool for millions of users around the world.
Source: 9to5google.com/2025/04/07/google-maps-fake-reviews/
Microsoft celebrates 50 years – special wallpapers, themes and bonus content
To celebrate the 50th anniversary of its founding, Microsoft has prepared limited edition graphic themes and retro elements for Windows and Xbox users.
Microsoft was founded on 4 April 1975 and today marks its half-century anniversary. On a special Unlocked page, it has published a review of its history and offered fans the chance to download special anniversary wallpapers for Windows and Xbox that pay tribute to iconic games and apps like Solitaire and Sketchpad. In addition to the backgrounds, the reverse of the cards can also be changed in the Microsoft Solitaire Collection available in the Microsoft Store.
The Microsoft Edge browser has been given a special nostalgic anniversary theme that’s available exclusively for the browser. Plus, the Clipchamp video editor has added anniversary elements like retro icons, nostalgic windows, and high-resolution graphics that you can embed in your videos. So Microsoft is offering modern users the chance to reminisce about the old days with a modern design.
Source: unlocked.microsoft.com/50th/
Google is moving to a single Google.com domain
Google is unifying search under a single domain – google.com, ending the era of national domains such as google.sk or google.fr. The change is mainly technical and will not affect the location of search results.
Starting in April 2025, Google will globally redirect all of its national search engines to google.com. Although you will no longer see a domain like google.sk, results will still be tailored to your location. That’s because Google has been using the physical location of your device, not the domain, to determine your location since 2017. So for the user, the change is purely cosmetic – the only difference is that only google.com will be displayed in the browser. In some cases, you may have to re-set your language or region.
With this move, Google is simplifying the management of its global service, reducing technical complexity and better responding to an era where mobile devices and AI locate users much more accurately than URL domains.
Source: 9to5google.com/2025/04/15/google-com-search/
Snipping Tool for Snipping in Windows gets text recognition (OCR)
Windows 11 gets a handy enhancement – text recognition(OCR) directly in the Snipping Tool . The feature will allow you to extract text from images without the need to use external tools. Microsoft is adding a new text extraction button to the tool that will make OCR mode available. The user will be able to select specific text using the cursor or mark the entire area, and the application will automatically recognize and copy the text. There will also be options to remove line breaks, which will make subsequent editing easier in Notepad or Word, for example.
The new feature is only available for Insiders in version 11.2503.27.0 for now. You open the tool as always via the Win + Shift + S shortcut, while OCR mode allows you to extract text quickly and without having to save the image to disk.
Source: blogs.windows.com/windows-insider/2025/04/15/text-extractor-in-snipping-tool-begins-rolling-out-to-windows-insiders/
Meta AI has officially arrived in Slovakia and the Czech Republic
Meta has finally launched its Meta AI for users in Slovakia and the Czech Republic, expanding the tool’s availability across Europe.
You can try out Meta AI via Meta.ai, and you may need to log in via Facebook or Instagram to access it. The tool offers features such as generating images, answering questions or helping with various tasks, making it a competitor to well-known solutions such as ChatGPT, Copilot or Gemini. Although the service is available in a limited mode for now, it is the next step in the development of AI tools available to ordinary users.
Source: meta.ai/
YouTube celebrates 20 years – the first video is still online
23.4.2025 marks exactly 20 years since the first video was uploaded to YouTube. The platform that fundamentally changed the way we consume content online is marking this milestone with special features and a return to its roots.
The first YouTube video was Me at the zoo, uploaded by co-founder Jawed Karim on April 23, 2005. The video is only 18 seconds long and shows him at the San Diego Zoo talking about elephants. Although YouTube today has videos in 4K, 8K or VR format, this 240p pre-flood video remains a symbol of the platform’s early days. In 2025, it already has more than 355 million views and 17 million likes. To mark the anniversary, YouTube has added a digital cake and an elephant as a tribute to its history.
Today, YouTube has 2.7 billion active users who watch videos at least once a week. On top of that, there are tens of millions of video creators who regularly feed YouTube. Over three million channels are signed up for the affiliate program and Google shares ad revenue and premium subscriptions with them. In the last three years alone, it has paid them over $70 billion.
Source: androidauthority.com/youtube-first-video-upload-20th-anniversary-3547053/
March 2025
Firefox extends support for Windows 7 until 2025
Mozilla has decided to extend support for the legacy Windows 7 operating system through a special version of Firefox – ESR 115. This extended support version will receive security updates until at least September 16, 2025, and a decision will be made in August this year on a possible further extension. This is good news for Windows 7 users, as most browsers have already discontinued support for this system. The ESR version of Firefox will allow you to continue using the browser securely, but without the latest feature updates. This model is especially beneficial for businesses and those who prefer stability over frequent software changes.
Source: whattrainisitnow.com/release/?version=esr
FreeTube brings ad-free and private YouTube
The FreeTube app offers you a way to watch YouTube videos without ads and without having to sign in. It works on a variety of platforms, including ARM devices, and offers a portable version that you can run directly from a flash drive. The user interface is reminiscent of YouTube, but it’s minimalistic and free of distractions. You can play videos directly or download them and watch them offline. FreeTube allows you to create playlists and watch channels without linking to a Google account, increasing your level of privacy.
Source: github.com/FreeTubeApp/FreeTube
Google Gemini enables free analysis of (even scanned) documents
Google Gemini now makes it possible to analyze a variety of documents without having to pay, making this feature available to a wider audience. Users can upload images, PDFs, and Word documents directly through the web interface or mobile app. However, paid accounts still offer more options – such as support for Excel spreadsheets or a larger contextual window.
The Gemini 2.0 Flash model is also suitable for programming, but the accuracy of the answers depends on a clearly defined task. The AI tool Grok from X (formerly Twitter) is often mentioned in this context, which produces different results depending on the accuracy of the requirements.
Source: x.com/GeminiApp/status/1892678157406281931
Antropic releases Claude 3.7 – the most intelligent AI available for free
A new version of Claude 3.7 Sonnet, which is significantly more powerful than the previous version 3.5, offers advanced reasoning and faster answers to users of the free version, while subscribers gain the ability to adjust the accuracy and length of answers up to 128,000 tokens.
In benchmarks, this model outperforms competitors such as DeepSeek and OpenAI o1 by more than 12%. Anthropic is also developing a Claude Code model for analysis and code generation, which is currently only available in a test version and is not free.
Source: anthropic.com/news/claude-3-7-sonnet
Microsoft tests ads in Office apps
Microsoft is experimenting with displaying ads in the free version of its office suite, Microsoft 365 Copilot. The ads appear on the right-hand side of Word, Excel, and PowerPoint and can’t be removed unless you subscribe to the paid version as a user.
So far, it’s just a testing, and the ads mainly promote Microsoft products such as Designer. Documents in this version of Office can only be stored on OneDrive, and free users have 5GB of storage. Despite the ads and restrictions, users can still edit documents after the trial period has expired, which was not possible before.
Source: beebom.com/microsoft-free-ad-supported-office-quietly-launched/
Copilot offers AI conversation on mobile for free
Microsoft Copilot is the latest to offer free access to voice interaction and “Think Deeper” features that were previously only available to paying users. As a user, you can have a natural conversation with the AI without having to reactivate it. This functionality is already familiar from ChatGPT, but in the case of Copilot, all you need to do is download the app and start using the AI to search for information or learn languages.
Competitor Perplexity offers a similar “Deep Research” feature for in-depth analysis. The paid version of Copilot Pro still comes with benefits such as higher query limits or image generation, but many features that were until recently chargeable are now available for free.
Source: microsoft.com/en-us/microsoft-copilot/blog/2025/02/25/announcing-free-unlimited-access-to-think-deeper-and-voice/
Microsoft unveils Majorana 1 quantum chip
Microsoft has unveiled its first quantum chip, called Majorana 1, which marks an important step in quantum computing. This chip is designed to be scalable up to a million qubits, which could significantly change the field of computing in the future. Quantum computers can perform calculations much faster than traditional computers and could bring breakthroughs in modeling chemical reactions, developing new materials, or finding ecological solutions.
Although fully functional quantum computers are still a thing of the future, Microsoft’s development has focused on the stability of qubits using a so-called topological conductor, a special material that allows the topological state of matter to be created. This should make quantum computing more stable and powerful. Quantum computers may have future applications in a variety of fields, including healthcare, agriculture, and ecology, and may help, for example, in the development of catalysts for the decomposition of plastic waste or the production of regenerative materials.
Source: news.microsoft.com/source/features/innovation/microsofts-majorana-1-chip-carves-new-path-for-quantum-computing/
Microsoft plans to end Skype after more than 21 years
According to the latest beta version of Skype for Windows, Microsoft plans to discontinue this popular communication service in May this year. Users are advised to switch to the free version of Microsoft Teams, which has gradually taken over the functions of Skype.
Skype has been the main tool for video calls over the internet for many years, but in recent years it has been overshadowed by modern applications such as WhatsApp, Zoom, and Microsoft Teams. Despite Microsoft’s constant improvements to Skype, its future has been uncertain since Teams’ launch in 2017. Although the company hasn’t officially announced the end of Skype yet, source code information suggests that its end is near.
Source: microsoft.com/en-us/microsoft-365/blog/2025/02/28/the-next-chapter-moving-from-skype-to-microsoft-teams/
Neural AI rendering coming to DirectX
Microsoft is adding Cooperative Vectors technology to DirectX to enable AI integration directly into the graphics pipeline. This technology will be supported on GPUs from Nvidia, AMD, Intel and Qualcomm and will enable cross-platform neural rendering.
The news is closely related to the so-called Neural Shaders, which are a feature of the new generation of Nvidia Blackwell GPUs. It allows developers to combine traditional shader computations with AI inference on tensor cores without the need to switch the GPU to a special mode. Developers will be able to use AI models to more realistically simulate materials, lighting, or physical effects in games. Cooperative Neural Vectors support will appear in the DirectX SDK as early as April 2025, but we’ll have to wait longer to see it actually used in games.
Source: devblogs.microsoft.com/directx/enabling-neural-rendering-in-directx-cooperative-vector-support-coming-soon/
Windows 11 gets a new battery indicator with coloured statuses and percentages
Microsoft tweaks the battery indicator in Windows 11 to make it clearer and more practical. The new version will add colour differentiation for battery status and the ability to display percentages right next to the icon. The battery will have four colour modes: green (charging and good condition), yellow/orange (power saving mode at ≤20%), red (critically low battery), and standard white (normal condition). The battery percentage display will be optional and can be turned on in the settings. The change is only available for the Insider version of Windows 11 for now, but it’s likely to make its way to the stable version in a future update.
Source: blogs.windows.com/windows-insider/2025/02/28/announcing-windows-11-insider-preview-build-27802-canary-channel/
Google introduces the new Gemma 3
Gemma 3 is Google’s latest multimodal AI model that handles text, image, and video analysis. It has better math skills, structured responses, and supports over 140 languages.
Running on a PC is possible via Ollama.com, where versions (1B, 4B, 12B, 27B) are available depending on the performance of the device. The most powerful variant offers a contextual window of 128,000 tokens. Community versions optimized for specific tasks, such as programming assistance, are also being developed.
Source: developers.googleblog.com/en/introducing-gemma3/
Boston Dynamics’ Atlas robot mimics human movements
The latest video shows that the humanoid robot Atlas can already mimic human gait and movements. Boston Dynamics used motion tracking of real people to help the robot learn natural movement, which is more challenging than acrobatic stunts. Atlas is a prototype for now, but the technology is improving all the time and moving toward a commercial version. However, competition is growing – several startups are working on bipedal robots for warehouses and homes, and Nvidia has developed Jetson Thor, a specialized computer to control them.
Source: Boston Dynamics
February 2025
Vivaldi 7.1 – a browser with better personalisation and privacy protection
Vivaldi, the browser known for its focus on privacy and customisability, has released version 7.1, which brings new features and improvements. Highlights include a weather widget integrated into the browser interface, better Dashboard personalisation options, and simplified adding of pages to the Quick Menu. It also allows users to quickly share tabs between devices using the “Send Tab to Device” feature. To maintain its independence, Vivaldi has changed some of the default search engines, with Startpage, Ecosia, DuckDuckGo and Qwant among the supported ones. The new version is also more stable and can recover from unexpected crashes. Vivaldi 7.1 is available for Windows, macOS and Android, with the Windows version already ready for download.
Source: vivaldi.com/new/
YouTube bug brought unskippable ads even an hour long
A bug on YouTube is causing ads to run for over an hour without the option to skip them. According to YouTube, this is due to ad-blocking tools that cause ads to display incorrectly and may hide the skip button. Normally, non-skippable ads are up to 15 seconds long, or up to 60 seconds in the case of the smart TV app.
YouTube has long restricted ad-blockers, forcing users to either accept ads or subscribe to YouTube Premium. Those who continue to use blockers may face unexpected problems, such as extremely long ads or the inability to skip them. The era of convenient ad-free YouTube viewing with ad-blockers seems to be slowly coming to an end.
Source: androidauthority.com/youtube-long-unskippable-ads-problem-3519957/
Sony ends Blu-ray disc production after 18 years
Sony will close its last factory producing recordable Blu-ray discs, MiniDV tapes and MD Data media this month. However, Blu-ray technology is not completely finished – production of pre-recorded movie and game discs continues, and Sony will retain production lines for business partners as long as it is economically sustainable. Blu-ray originally gained traction thanks to the PlayStation 3 and the support of movie studios, but is now taking a back seat to the growing popularity of cloud services and digital distribution of games and movies.
However, optical discs still have advantages, especially for storing large amounts of data for long-term backup. Other manufacturers such as Panasonic, Pioneer, Verbatim and Ritek remain on the market and will continue to produce recordable discs, even if in more limited quantities. Physical media will not become completely obsolete, but their role in the consumer segment will continue to decline.
Source: tomshardware.com/pc-components/storage/after-18-years-blu-ray-media-production-draws-to-a-close-sony-shuts-its-last-factory-in-feb
Google Gemini 2.0 is coming – a new model focused on AI agents
Google has launched Gemini 2.0, the latest version of its AI model designed for the “agent era”. This model brings advanced multimodal capabilities, including native image and audio generation, as well as integration with tools such as Google Search and coding features. Gemini 2.0 Flash, an experimental version of the model, is already available to developers through Google AI Studio and Vertex AI, with wider availability planned for early next year.
In addition, Google is showcasing projects such as Astra and Mariner, which demonstrate the potential of real-time AI agent systems. The company highlights its commitment to responsible AI development, implementing safety measures and working with external experts to test and improve the model.
Source: blog.google/products/gemini/
EU regulates AI – new regulation brings bans and penalties
The European Union is responding to the risks associated with artificial intelligence (AI) with a new regulation that bans the use of AI in areas considered high-risk. The Artificial Intelligence Act (EU AI Act) came into force on 2 February 2025 and imposes strict rules on the development and use of AI in the EU.
The ban covers manipulative AI systems, social assessment of citizens, non-targeted collection of CCTV data, and biometric analysis in public places. AI systems must also not be used to monitor emotions in workplaces or schools. The regulation also applies to large AI models (e.g., IBM Granite, Meta Llama 3) and sets high fines for breaches – up to 7% of a company’s global annual turnover. Personal and scientific AI systems are exempt.
Source: ibm.com/think/topics/eu-ai-act
Meta faces accusations of using pirated books to train AI
Meta is facing accusations of copyright infringement after it emerged that it used pirated books downloaded from torrents to train its LLaMA AI models. Internal emails uncovered during the lawsuit indicate that the company downloaded approximately 81.7 terabytes of data from shadow libraries, such as LibGen and Z-Library, through a site called Anna’s Archive. Some employees raised concerns about the ethics and legality of such actions.
Meta defends itself by claiming that, although it downloaded the material, it took measures to prevent its further distribution (seeding) during torrenting. The company also argues that the use of this data for AI training falls under ‘fair use’. However, the authors, including Ta-Nehisi Coates and Sarah Silverman, claim that Meta knowingly used pirated versions of their works without permission, which constitutes copyright infringement. This case highlights the growing tension between content creators and tech giants over the use of protected materials to train AI systems.
Sources: arstechnica.com/tech-policy/2025/02/meta-torrented-over-81-7tb-of-pirated-books-to-train-ai-authors-say/ and arstechnica.com/tech-policy/2025/02/meta-defends-its-vast-book-torrenting-were-just-a-leech-no-proof-of-seeding/
Google appreciates Chinese AI. DeepSeek impressed with efficiency, but didn’t bring anything groundbreaking
Google DeepMind’s CEO, Demis Hassabis, called DeepSeek’s AI model “probably the best work out of China”. However, despite the overwhelming response, he stressed that the model does not represent a major scientific advance, as it uses techniques already known in the field of AI. DeepSeek impressed by claiming that it trained its model at a significantly lower cost and on less powerful Nvidia chips, sparking debates about the efficiency of spending by large AI companies.
However, Hassabis cautioned that these costs may actually be higher than the company is reporting. He also noted that Google’s Gemini 2.0 Flash models are more efficient. The artificial general intelligence (AGI) debate is gaining momentum, with Hassabis estimating its arrival within five years, which will require the company to prepare thoroughly for its implications.
Source: cnbc.com/2025/02/09/deepseeks-ai-model-the-best-work-out-of-china-google-deepmind-ceo.html
Google Chrome could automatically change our passwords for us in the future
Google is working on a new feature for Chrome that could significantly improve user security. This involves automatically changing passwords when a data leak is detected. Chrome would track information about security incidents and offer an instant password change option in the event of an account breach. This process would be fully automated and would complement Google’s current password manager. Users wouldn’t have to change their passwords manually – Chrome would take care of it for them. The new feature is still under development, and it is not known when it will be available to the public.
Source: androidheadlines.com/2025/02/google-chrome-change-passwords-automatically.html
OpenAI is preparing new models – GPT-4.5 Orion coming soon, GPT-5 later
OpenAI is preparing new versions of its language models. Sam Altman announced that GPT-4.5 Orion will be coming in a few weeks and will be the last model that does not take the previous context into account.
GPT-5 will bring significant enhancements such as an expanded context window and multimodal capabilities, allowing you to work with text, images, and audio. It is expected in a few months. OpenAI also plans to merge models with the “o” attribute, and non-paying users will gain access to GPT-5, although with limited features. The paid Pro version will offer enhanced capabilities such as unlimited voice interaction and advanced features in the Sora video creation tool.
Zdroj: com/sama/status/1889755723078443244
Windows 11 wil upgrade Snap Panel – a more convenient window layout
Microsoft is planning improvements to the “Snap” feature in Windows 11, which allows you to quickly arrange windows on the screen. The new floating bar visual will be clearer, and Microsoft wants to encourage the use of the Win+Z keyboard shortcut, which doesn’t work yet. The changes appeared in Insider Preview Build 22635.4805. Window snapping has evolved from a simple feature in Windows 7 to the current half- or quarter-screen customization options. Users can arrange windows not only with the mouse, but also with the Win + left/right arrow keys shortcut for more efficient multitasking.
Source: blogs.windows.com/windows-insider/2025/01/24/announcing-windows-11-insider-preview-build-22635-4805-beta-channel/
Tomb Raider IV-V-VI Remastered – Lara Croft returns in a new guise
Remastered versions of the next three Tomb Raider episodes are here, bringing improved graphics, new features, and fidelity to the original gameplay. Players can switch between the old and new visuals with a single press of the F1 key.
The games retain the classic controls, but also offer a modernized scheme for more comfortable gameplay. From the beginning, the developers at Aspyr Media planned a remaster of the entire six-part series from Core Design. Improvements include redesigned lighting, better object visibility, and the ability to display boss health. Players can also look forward to the “Times Exclusive” bonus level, which many fans of the original games haven’t played.
Source: rockpapershotgun.com/tomb-raider-iv-vi-remastered-review
Kingdom Come: Deliverance II – the best historical RPG?
Kingdom Come: Deliverance II is the sequel to the successful historical role-playing game from Czech studio Warhorse Studios, released on 4 February 2025 for PlayStation 5, Windows, and Xbox Series X/S. The game is set in 15th-century medieval Bohemia and follows the story of Henry, the son of a blacksmith from Skalica, who once again embarks on an epic adventure just two days after the end of the first instalment.
The game has received mostly positive reviews from critics, with an average score of 88% on Metacritic and even 92% from over 40,000 players on Steam. Reviewers praise the game’s realistic depiction of the medieval world, detailed environments, and deep story. For example, the Dexerto server gave the game a full 5/5 rating, praising the gorgeous open world and the improved combat system. Similarly, Windows Central awarded 5/5, highlighting the excellent combat and unexpected twists in the story.
The game was also commercially successful, selling more than a million copies on its first day and over 2.5 million copies within three weeks. This success confirms players’ high interest in quality historical RPG titles.
Kingdom Come: Deliverance II offers a vast open world that faithfully recreates the Czech landscape, including areas like Český ráj, Trosky Castle, and Kutná Hora. Its focus on realism—requiring players to manage sleep, diet, and hygiene—deepens immersion into medieval life.
Regarded as a milestone in the RPG genre, the game blends historical accuracy with compelling storytelling and complex mechanics.
Source : indian-tv.com/reviews/other-dil-is-the-best-thing-to-kingdom-come-deliverance-kvv307
January 2025
Free web search with ChatGPT for all users
ChatGPT now allows all users to search the web, greatly enhancing its ability to provide comprehensive answers. The feature, which was previously only available to ChatGPT Plus subscribers or users with priority access, now allows the chatbot to search for information on websites, social networking sites or the YouTube platform. Users can verify these sources through links provided by the chatbot. This innovation is particularly useful in an academic environment, where it makes it easier for students to access verified resources.
The feature is available on both the web and mobile interfaces and can be activated by clicking on the globe icon next to the search bar. ChatGPT thus offers new ways of working with information, regardless of whether the user has a subscription or not.
Source : theverge.com/2024/12/16/24322665/chatgpt-search-engine-rolling-out-free-users
Elon Musk prepares Xmail – a new email client with interesting features
Elon Musk, known for his ambitious technology projects, is planning to unveil a new email client called Xmail. The aim is to provide a simple and friendly environment that could compete with Google’s dominant Gmail. Xmail was mentioned on Platform X (formerly Twitter), where Musk responded to a user request for a less formal and complicated email system. Although specific details and a launch date have yet to be announced, Xmail is expected to offer unique features that could attract users.
Xmail’s success will depend on its ability to deliver revolutionary innovations that overcome long-standing Gmail user habits. One tempting feature could be the ability to use shorter email addresses in the form of user@x.com, which could appeal to technology enthusiasts. Musk’s ambition to dethrone Google shows that he has big plans for email too.
Source: androidheadlines.com/2024/12/elon-musk-hints-at-working-on-xmail-a-new-dm-style-email-service.html
Qualcomm wins dispute with ARM
Qualcomm has won an important victory in a legal battle with ARM over licensing agreements following its acquisition of startup Nuvia in 2021. A US federal court ruled that Qualcomm had not breached the licensing agreements, allowing it to continue developing and selling chips based on Nuvia’s technology. Despite the ruling, the judge warned that the verdict was not conclusive and that the question of ARM’s royalty and legal claims remained open.
This controversy is significant for the future of Qualcomm’s chips, particularly its new flagship Snapdragon 8 Elite with Oryon cores, which uses Nuvia’s Phoenix architecture. This allows Qualcomm to continue its expansion in the AI-focused computer chip market, but the threat of further legal action from ARM could affect the long-term development of the technology.
Source: reuters.com/legal/us-jury-deadlocked-arm-trial-against-qualcomm-still-deliberating-2024-12-20/
Instagram launches long-awaited message scheduling feature
Instagram launches long-awaited message scheduling feature Instagram, which has evolved from a photo-sharing platform to a full-blown social network, is finally introducing a feature that users have been asking for for years – the ability to schedule messages. Users can now write a message and schedule it to be sent up to 29 days in advance. This feature is perfect for those who don’t want to send messages at inconvenient times, or for those who want to set reminders without worrying about forgetting.
Message scheduling is available on both Android devices and iPhones. Simply press and hold the send button, select the date and time from the scroll bar and the message is ready to send. Users can edit or cancel a scheduled message at any time via a notification in the conversation. As part of Instagram’s suite of New Year’s Eve features, this new feature makes it easy to prepare holiday wishes or other important messages in advance without stress.
Source: businesstoday.in/technology/news/story/instagram-launches-dm-scheduling-feature-letting-users-plan-messages-up-to-29-days-ahead-457687-2024-12-18
USB-C connector made mandatory by the European Union
The European Union has introduced EU Directive 2022/2380, which requires all new devices on the EU market to use a USB-C charging port from 28 December 2024. This decision aims to simplify the use of chargers and reduce e-waste. The regulation applies not only to smartphones, but also to tablets, e-book readers, computer peripherals and headphones. Most smartphones already use USB-C and Apple has adapted its new iPhones to the standard. Laptops have a longer transition period and will not need to support charging through the USB-C port until 28 April 2026. Despite the widespread adoption of USB-C, questions remain about the compatibility and performance of chargers, as new devices often support super-fast charging, which may require the purchase of additional accessories. However, the move is seen as a significant shift towards a greener and more user-friendly technology environment.
Source: commission.europa.eu/news/eu-common-charger-rules-power-all-your-devices-single-charger-2024-12-28_en
OpenAI introduces a new o3 model for ChatGPT
ChatGPT will launch a new model called o3, which is expected to reach the level of Artificial General Intelligence (AGI), meaning that it will surpass human capabilities in many aspects. The model will be available in several versions, including o3 and o3-mini. The move comes as a successor to the o1 models, omitting the o2 version to avoid trademark disputes with UK operator O2. The model is expected to be launched at the end of January 2025.
The new model focuses on accuracy and justification of results, particularly in maths, physics and science. Interestingly, tests have shown that the o3 model tends to be more convincing at lying (hallucinating), which represents a form of optimization in an attempt to strike a balance between efficiency and accuracy of answers. The o3 version will also allow users to adjust the length of time the model spends ‘reasoning’, thus adapting to different needs and requirements.
Source: techcrunch.com/2024/12/20/openai-announces-new-o3-model/
Facebook stops fact-checking. Freedom of speech or the road to misinformation?
Facebook (Meta) has announced that it is ending its fact-checking program, which verified the accuracy of posts on the platform. According to Mark Zuckerberg, the move is intended to promote freedom of expression, as the social network’s algorithms often mistakenly censor legitimate posts. Fact-checking will be replaced by ‘community notes’, similar to those used by X (formerly Twitter). The withdrawal of cooperation with journalists and organizations such as PolitiFact raises fears of an increase in misinformation and fraud on the platform.
The change comes against a backdrop of past criticism, particularly from Donald Trump, who has accused Facebook of interfering in elections. Experts warn that the decision could lead to an even greater spread of fake news and propaganda, which could impact elections and global events. As the fact-checking program ends in the US, a similar change is expected to gradually spread to other countries, including Slovakia.
Source: tampabay.com/news/florida-politics/2025/01/07/meta-ending-third-party-fact-checking-partnership-with-us-partners/
Microsoft Office ends with updates to Windows 10
Microsoft has announced that support for Microsoft 365 apps on the Windows 10 operating system will end on 14 October 2025. This date also marks the end of official support for Windows 10. After this date, both the Office apps and the system itself will still work, but they will no longer receive updates, which could make them more vulnerable to security threats over time.
Microsoft is also urging users to migrate to Windows 11, which is considered a stable and secure platform. However, some users may experience compatibility issues with their devices, mainly due to processor and TPM 2.0 chip requirements. While it is currently possible to work around these restrictions, the company plans to tighten the rules in the future.
Source: support.microsoft.com/en-us/office/what-windows-end-of-support-means-for-office-and-microsoft-365-34e28be4-1e4f-4928-b210-3f45d8215595
The PC market grew only moderately in 2024
In 2024, the PC industry grew slightly by 1.3%, with total shipments reaching 245.3 million units. Despite manufacturers’ efforts to push new technologies such as artificial intelligence and Windows 11, the market remains overshadowed by the post-pandemic downturn. The biggest winners were Asus, Lenovo, and Apple, which saw the most significant gains in market share. Conversely, HP and Dell experienced slight declines.
Although marketing around AI-enabled laptops and the Snapdragon platform has made progress, consumer interest in these features has been lukewarm. Microsoft and Google are trying to increase appeal by integrating AI into software, but they have not made significant progress. As a result, the PC market is not expected to gain momentum until 2025, when innovation and the opening of Windows on Arm to new chip platforms could provide the expected boost.
Source: gartner.com/en/newsroom/press-releases/2025-01-15-gartner-says-worldwide-pc-shipments-increased-1-point-4-percent-in-fourth-quarter-of-2024
Android 16 brings new multitasking features to tablets
Google is preparing Android 16 with Baklava, which is expected to be released in the second quarter of 2025 and will reach platform stability as early as March. A major innovation will be improved multitasking for tablets, allowing up to three apps to be displayed simultaneously on a single screen. This feature will significantly improve the use of large screens, which until now could only be fully exploited by devices with their own add-ons, such as Samsung Galaxy tablets.
The new flexible multitasking system, already available in Android 16 Developer Preview 2, is designed to make it easier to view apps on larger screens. The Android 16 beta program, scheduled for later this month, could reveal more details. The feature is aimed at increasing productivity and convenience for tablet and foldable device users.
Source: 9to5google.com/2025/01/16/android-16-split-screen-mode-tablets-upgrade-rumor/
TSMC opens its first semiconductor factory in the USA
Taiwan-based TSMC, a leader in semiconductor manufacturing, is opening its first factory in the US, specifically in Arizona, with production scheduled to begin in 2025. The fab will use 4 nm technology and is currently in pilot production. In 2028, TSMC plans to open a second fab for more advanced 3 nm and 2 nm processes, with the possibility of further expansion by the end of the decade. Key customers will be companies such as Nvidia and Apple.
Despite the expansion, most production, including high-end technology, remains in Taiwan. TSMC CEO C.C. Wei pointed to the challenges and higher costs associated with US production, with a total investment of $65 billion for the three US plants. The move is a significant contribution to diversifying and securing global supply chains in the semiconductor industry.
Source: reuters.com/technology/tsmcs-us-plant-unlikely-get-latest-chip-tech-before-taiwan-ceo-says-2025-01-17/
35 years since the first ransomware
At the turn of 1989 and 1990, the first ransomware, called AIDS, appeared. This malware was physically distributed by mail on 5.25-inch floppy disks and targeted DOS-based PC users. AIDS encrypted file names after 90 computer boots and demanded a ransom of $189 or $378 to unlock them, to be sent to a post office box in Panama.
The ransomware was stopped in January 1990 when expert Jim Bates created a tool to remove it. The author was Dr Joseph Popp, an evolutionary biologist involved in AIDS research. He was charged with extortion, but his mental health landed him in a psychiatric ward rather than prison. This event is now seen as a historic milestone in the field of cyber threats.
Source: edition.cnn.com/2021/05/16/tech/ransomware-joseph-popp/index.html
- The best IT podcasts: both interesting and entertaining. Which podcast is worth listening to?
- The latest IT news from the tech world
- Book Review of Head First Java (3rd Edition)
- German language level on the CV. Kedy si napísať úroveň B2 či C1?
- Java vs C# – comparison of programming languages (2024)
- The most used AI browsers and tutorials on how to work with AI chatbots
