Pissed-off Fans Flooded the Twin Peaks Reddit With AI Slop To Protest Its AI Policies

Moderators reversed course on its open door AI policy after fans filled the subreddit with AI-generated Dale Cooper slop.

The SIM Farm Hardware Seized by the Secret Service Is Also Popular With Ticket Scalpers

The tech the Secret Service claims can be used to "disable cell phone towers" is very commonly used by ticket scalpers to game Ticketmaster.

How Surveillance Firms Use ‘Democracy’ As a Cover for Serving ICE and Trump

Multiple Palantir and Flock sources say the companies are spinning a commitment to "democracy" to absolve them of responsibility. "In my eyes, it is the classic double speak," one said.

Microsoft teams up with Anthropic AI for 365 Copilot

The rollout starts today for Microsoft 365 Copilot customers in the Frontier program and Copilot Studio users.

Microsoft users get more AI model choice in Copilot

The tech giant goes multi-model with the addition of generative AI models from an OpenAI rival.

Five AI Data Centers Confirmed as Part of Stargate Project

OpenAI plans five new US data centers with Oracle and SoftBank as part of Trump's Stargate project, adding 7 gigawatts of AI computing capacity.

MLB to Roll Out AI-Powered Challenge System in 2026

The new system, which uses Hawkeye, is pitched as a compromise between robotic and traditional umpires

Oracle's $300B wager on OpenAI could reshape tech landscape

How long the generative AI boom will last, whether OpenAI can penetrate the enterprise market and if Oracle can scale up fast enough will likely decide if the pact flourishes or flounders.

Samsung benchmarks real productivity of enterprise AI models

Samsung is overcoming limitations of existing benchmarks to better assess the real-world productivity of AI models in enterprise settings. The new system, developed by Samsung Research and named TRUEBench, aims to address the growing disparity between theoretical AI performance and its actual utility in the workplace. As businesses worldwide accelerate their adoption of large language […]

The post Samsung benchmarks real productivity of enterprise AI models appeared first on AI News.

Inside Huawei’s plan to make thousands of AI chips think like one computer

Imagine connecting thousands of powerful AI chips scattered in dozens of server cabinets and making them work together as if they were a single, massive computer. That is exactly what Huawei demonstrated at HUAWEI CONNECT 2025, where the company unveiled a breakthrough in AI infrastructure architecture that could reshape how the world builds and scales […]

The post Inside Huawei’s plan to make thousands of AI chips think like one computer appeared first on AI News.

Baidu’s Apollo Go Begins Autonomous Driving Trials on Dubai Roads

The company has launched a 50-vehicle fleet following receipt of approval from the RTA in Dubai.

The post Baidu’s Apollo Go Begins Autonomous Driving Trials on Dubai Roads appeared first on Analytics India Magazine.

Microsoft Develops New AI Chip Cooling System, says It’s 3x Better Than Current Systems

“If you’re still relying heavily on traditional cold plate technology, you’re stuck.”

The post Microsoft Develops New AI Chip Cooling System, says It’s 3x Better Than Current Systems appeared first on Analytics India Magazine.

Databricks and OpenAI Strike $100M Deal to Bring GPT-5 to Enterprises

The partnership aims to bring GPT-5 and frontier AI models natively to 20,000+ enterprises.

The post Databricks and OpenAI Strike $100M Deal to Bring GPT-5 to Enterprises appeared first on Analytics India Magazine.

Indian College Placements are a Joke

“You either get a good referral, or you rot. Skills don’t matter unless you can show them in the exact format companies want.”

The post Indian College Placements are a Joke appeared first on Analytics India Magazine.

Salesforce Confident in India’s Growth Amid Push for Local Solutions

“It’s important to experiment and learn. Every failure teaches us how to do things better,” Salesforce CEO said. 

The post Salesforce Confident in India’s Growth Amid Push for Local Solutions appeared first on Analytics India Magazine.

Mangaluru Looks to Build Its Own Tech Identity, Not Replicate Bangalore

“The coastal city could showcase tangible results by applying deep tech to areas it already dominates”

The post Mangaluru Looks to Build Its Own Tech Identity, Not Replicate Bangalore appeared first on Analytics India Magazine.

NxtGen’s Standardised AI Solutions Framework Sets the Pace for AI Delivery at Scale

The framework, SAS-F, outlines an approach that involves real-time data ingestion, foundation model fine-tuning and agentic workflows.

The post NxtGen’s Standardised AI Solutions Framework Sets the Pace for AI Delivery at Scale appeared first on Analytics India Magazine.

Senate staff probes DOGE, finds locked doors and windows covered with trash bags

Democratic report describes Social Security risk and secretive DOGE offices.

Sierra’s Dream Chaser is starting to resemble a nightmare

"Development of new space transportation systems is difficult."

Felony charges after South Carolina high school filled with “fart spray”… for weeks

Let's dig into the science of stink.

Amazon Fire TV devices expected to ditch Android for Linux in 2025

Ditching Google could help Amazon better monetize its streaming gadgets.

High above the equator, Russia is stalking satellites used by NATO armed forces

"Russia and China already occupy important strategic hills and mountains in space."

ChatGPT Pulse delivers morning updates based on your chat history

New mobile chatbot feature analyzes conversations overnight.

Study: Planned budget cuts would hurt drug development badly

Drug patents frequently cite research that the NIH wouldn't be able to fund.

Experts urge caution about using ChatGPT to pick stocks

AI-selected portfolios might perform well in a growing market, but experts warn of downturn risks.

Amazon blamed AI for layoffs, then hired cheap H1-B workers, senators allege

Tech firms pressed to explain if H-1B workers are paid less than US workers.

Apple iPhone 17 review: Sometimes boring is best

The least exciting iPhone this year is also the best value for the money.

Amazon agrees to make canceling Prime easy, will refund customers $1.5B

Amazon’s settlement with FTC kills the “No, I don’t want free shipping” button.

Google DeepMind unveils its first “thinking” robotics AI

DeepMind researchers believe this is the dawn of agentic robots.

Reviewing iOS 26 for power users: Reminders, Preview, and more

These features try to turn iPhones into more powerful work and organization tools.

Apple demands EU repeal the Digital Markets Act

"DMA compliance is not optional, it's an obligation," the EU reminds Apple.

Astra’s Chris Kemp woke up one recent morning and chose violence

"We bought the engine from them, and it was garbage."

As many as 2 million Cisco devices affected by actively exploited 0-day

Search shows 2 million vulnerable Cisco SNMP interfaces exposed to the Internet.

Ford decides to run its Le Mans program in-house, racing in 2027

Instead of contracting an outside team to campaign it, Ford Racing gets the job.

DeepMind’s robotic ballet: An AI for coordinating manufacturing robots

An AI figures out how robots can get jobs done without getting in each other's way.

AI systems can easily lie and deceive us – a fact researchers are painfully aware of

In stress-testing AI models, it’s not hard to push them to the brink and make them threaten to harm humans.

Zelensky says a destructive drone arms race looms – but dystopia isn’t inevitable

The new age of drones and AI may be terrifying, but there are reasons for thinking the dystopian possibilities of future war will be contained.

AI in Africa: 5 issues that must be tackled for digital equality

Policy-makers must grapple with an uncomfortable truth: without deliberate action, AI will magnify global divides.

4 films that show how humans can fortify – or botch – their relationship with AI

From 1980s sci-fi films to today’s blockbusters, filmmakers have wrestled with questions about what happens when humans rely on intelligent machines.

‘Digital brains’ that ‘think’ and ‘feel’: why do we personify AI models, and are these metaphors actually helpful?

Metaphors help us understand complexity, but they can also be dangerous, or even misleading.

From South Park v Trump to AI slopaganda: deepfakes are now part of the news cycle, for better and for worse | Anna Broinowski

Deepfakes come with risks that demand urgent regulation. But it is vital their potential as a creative and satirical tool isn’t stifled

Salman Rushdie believes AI will not be a threat to authors until ChatGPT can write “a funny book”. His faith in human over synthetic creativity may hold some truth in the literary space. But on our screens – from film, art and satire to the algorithmically turbo-charged, factually opaque, monetised churn of the 24/7 news cycle – AI is already making us laugh.

Deepfakes – synthetic audio and video of people doing and saying things they never said or did – are the chief comedic disruptors in a suite of increasingly persuasive AI tools shaping the post-truth reality envisioned by the Microsoft engineer Eric Horvitz, where fact and fiction are indistinguishable. In eight short years, deepfakes have risen from cultural outlier to mainstream meme, embodying the futurist Roy Amara’s Law: we overestimate the effects of new technology in the short run but underestimate its long-term impacts.

Continue reading...

Elon Musk’s xAI accuses OpenAI of stealing trade secrets in new lawsuit

Suit alleges OpenAI has a ‘troubling pattern’ of hiring former xAI workers to access secrets about the Grok chatbot

Elon Musk’s artificial intelligence startup xAI has accused rival OpenAI of stealing its trade secrets in a new lawsuit, the latest in Musk’s legal assault on his former business partner, Sam Altman.

The lawsuit, filed on Wednesday in California federal court, alleged that OpenAI was engaged in a “deeply troubling pattern” of hiring away former xAI employees to gain access to trade secrets related to its AI chatbot Grok. The company says OpenAI is pursuing unfair advantages in the race to develop AI technology.

Continue reading...

Spotify removes 75m spam tracks in past year as AI increases ability to make fake music

Streamer to crack down on AI-generated spam by introducing filter to identify fraudulent uploads

Spotify has revealed it removed 75m spam tracks from its platform over the past year as artificial intelligence tools increase the ability of fraudsters to create fake music.

The world’s biggest music streaming service announced a crackdown on vexatious tracks after admitting the rise of powerful AI tools had coincided with a significant amount of spam content being tackled by the streamer.

Continue reading...

Researchers Just Found Something Extremely Alarming About AI’s Power Usage

It's even worse than we thought.

The post Researchers Just Found Something Extremely Alarming About AI’s Power Usage appeared first on Futurism.

Fiverr Told Its Employees to Embrace AI, Then Stabbed Them in the Back

Where are fired Fiverr employees going to get work? On Fiverr?

The post Fiverr Told Its Employees to Embrace AI, Then Stabbed Them in the Back appeared first on Futurism.

New AI Can Control a Robot Even If Its Legs Get Chainsawed Off

"We built a robot brain that nothing can stop."

The post New AI Can Control a Robot Even If Its Legs Get Chainsawed Off appeared first on Futurism.

Psychiatric Facilities Are Being Bombarded by AI Users

"We are witnessing the emergence of an entirely new frontier of mental health crises."

The post Psychiatric Facilities Are Being Bombarded by AI Users appeared first on Futurism.

Google’s Conversational Photo Editor Is the Rare AI Feature People Will Actually Use

Google’s tool greatly simplifies photo editing; just tell your phone what changes you want in the photo, and it’ll execute them. It also hints at the coming leap in how we interact with computers.

Caltech’s massive 6,100-qubit array brings the quantum future closer

Caltech scientists have built a record-breaking array of 6,100 neutral-atom qubits, a critical step toward powerful error-corrected quantum computers. The qubits maintained long-lasting superposition and exceptional accuracy, even while being moved within the array. This balance of scale and stability points toward the next milestone: linking qubits through entanglement to unlock true quantum computation.

Trump Signs TikTok Order With US App Valued at $14 Billion

President Donald Trump advanced plans for American investors to buy TikTok’s US operations from its Chinese owner ByteDance Ltd., with officials setting a potential value of $14 billion and outlining measures to ensure security of the new venture.

Google-Linked Crypto Miner Plans $3 Billion Debt for Data Center

TeraWulf Inc. is expected to raise approximately $3 billion to support the build-out of its data centers via a structure supported by Google Inc., according to Patrick Fleury, the crypto miner’s finance chief.

Why Apple Still Hasn’t Cracked AI

The company has long been the king of consumer gadgets. Bloomberg Originals shows how its shortcomings in artificial intelligence could put that at risk.

Intel Approaches Apple About Investment | Bloomberg Tech 9/25/2025

Bloomberg’s Caroline Hyde and Ed Ludlow discuss reports that Intel approached Apple about a possible investment and are working together more closely. Plus, HSBC says it has achieved a world-first in deploying quantum computing in financial markets, using IBM's Heron quantum processor. And Disney is gearing up for a legal fight with President Donald Trump over the reinstatement of Jimmy Kimmel's late-night show. (Source: Bloomberg)

CoreWeave Expands OpenAI Deals to as Much as $22.4 Billion

CoreWeave Inc. expanded its agreements to supply data center capacity to OpenAI by as much as $6.5 billion to $22.4 billion, the latest deal to underscore the immense demand for AI computing power.

CEO of Controversial Startup Vows to Keep Mass Publishing AI Podcasts Despite Backlash

Jeanine Wright, the co-founder of Inception Point AI, says that within 12 to 24 months, it’s human-created content that will be the anomaly

Amazon to Pay $2.5 Billion Over Prime Subscriptions

Amazon.com Inc. agreed to pay $2.5 billion in penalties and refunds and change its process for how to cancel its Prime subscription to settle a lawsuit by the US Federal Trade Commission.

OpenAI Searches for Ways to Finance Trillion-Dollar AI Plan

Sam Altman and his partners intend to get creative to cover the growing cost of data centers and chips for AI. But first…

Oracle, Silver Lake, MGX Weigh 45% TikTok Stake, Board Seats

Oracle Corp., Silver Lake Management LLC, and the Abu Dhabi-based investment company MGX are in talks to invest in a US version of TikTok and receive board seats in the new venture, according to people familiar with the discussions.

Google Asks US Supreme Court to Pause App Store Overhaul

Alphabet Inc.’s Google asked the US Supreme Court to pause a lower court’s order requiring an overhaul of the technology giant’s app store policies in an antitrust case filed by Fortnite-maker Epic Games Inc.

Big Tech’s Costly AI Race Is Fueling Hundreds of Billions of Dollars in Debt Deals

Technology firms with vast funding needs to pay for their AI ambitions are striking blockbuster debt deals at the fastest pace in years, taking advantage of near—insatiable investor appetite to lock in financing for initiatives whose ultimate payoff remains uncertain.

Hate Groups Seize on Trump’s Antifa Order With Online Threats

Online extremists are rallying around President Donald Trump’s designation of Antifa as a domestic terrorist organization following conservative activist Charlie Kirk’s death, raising concerns that social media harassment could turn violent, according to researchers who track hate groups.

ChatGPT Feature Offers Users a Personalized Daily Briefing

OpenAI is rolling out a new ChatGPT feature that sends users a set of personalized news, research and other updates each day based on their prior conversations with the chatbot, an early attempt at making its flagship product more proactive.

Germany’s Industry Crisis Deepens as Bosch Cuts 13,000 Workers

Robert Bosch GmbH’s plan to slash 13,000 additional jobs shows how the German auto industry’s decline is rippling through Europe’s biggest economy.

Microsoft Disables Some Israeli Use on ‘Mass Surveillance’ Rules

Microsoft Corp., disabled some use of its software by the Israeli military after an investigation spurred by news reports that the company’s products were involved in the surveillance of civilians.

Meta Set to Face EU Finding It Failed to Police Illegal Posts

Meta Platforms Inc. is set to face a charge sheet from the European Union for failing to adequately police illegal content, risking fines for violating the bloc’s content moderation rulebook.

Constellation Software Tumbles as ‘Irreplaceable’ Founder Resigns

Mark Leonard resigned as president of Canada’s Constellation Software Inc. for health reasons, sending the shares to their lowest level in more than a year and prompting an analyst downgrade.

VW Loses Appeal Over Legality of Updated Diesel Emission Device

Volkswagen AG and Germany’s transport regulator KBA failed to overturn a lower court ruling that quashed a regulatory approval of a software device that deactivates emission cleaning in diesel engines depending on temperature.

Blackstone, Warburg Said to Vie for CVC-Backed Sebia Stake

Blackstone, Warburg Pincus and Nordic Capital are among private equity bidders looking to invest in French diagnostics provider Sebia, which could be valued at around €5 billion, according to people familiar with the matter.

Google Backs Crypto Miner Deal In Rush for AI Data Centers

Alphabet Inc.’s Google has agreed to anchor a $3 billion data center contract with Cipher Mining Inc. in the latest tie up between a crypto mining firm and a big tech company seeking computing resources for artificial intelligence.

Citigroup Moves Almost 1,000 Tech Jobs to India After China Cuts

Citigroup Inc. has moved close to 1,000 tech jobs to India’s business support centers following cuts to its workforce in China, according to people familiar with the matter.

Billionaire Xiaomi Founder Debuts a $630 Answer to iPhone 17

Xiaomi Corp.’s billionaire co-founder unveiled a $630 smartphone designed to take on the just-released iPhone 17, underscoring the Chinese company’s broader ambitions to take on US rivals from Apple Inc. to Tesla Inc.

Musk’s Grok AI Cleared for Use Across US Government Agencies

Elon Musk’s xAI signed a new agreement to expand access to its artificial intelligence chatbot Grok to the federal government, the General Services Administration announced Thursday.

Alphabet Is ‘Most Valuable Company’ on AI Strength, Moffett Says

Alphabet Inc. trails behind a few other megacap technology companies in size, but the Google parent may be destined to overtake them given its strong position in artificial intelligence and other key sectors, according to MoffettNathanson.

SAP Hit by EU Antitrust Probe Into Software Support Practice

SAP SE has been hit by a European Union antitrust probe into whether the German software giant distorted competition for on-premise maintenance and support services related to a management program it sells.

LG Electronics Said to Launch $1.3 Billion India IPO After Pause

LG Electronics Inc.’s India arm is set to launch its initial public offering to raise about 115 billion rupees ($1.3 billion) in the week beginning Oct. 6, according to people familiar with the matter.

The former Netflix exec who oversaw 'Stranger Things' and 'Bridgerton' is becoming Amazon's new TV head. Read the memo.

Peter Friedlander has been named head of TV at Amazon MGM Studios, replacing Vernon Sanders. He was a longtime Netflix exec.

We moved from Texas to California to be near my husband's family. I miss mine and sometimes wonder if we made the right choice.

My family moved from Texas to California. While there are advantages to being here, I'm not always sure we made the right choice.

Costco says new extended hours added 1% to weekly sales

Costco's CEO said the extended hours for executive members boosted weekly sales by 1% and led more US shoppers to upgrade.

Sam Altman explains why AI will replace 40% of your work

Sam Altman explains how AI is evolving, why regulation and safety matter, and what superintelligence could mean for the future of work and discovery.

Trump approved a $14 billion TikTok deal in an executive order, and employees have big questions

President Donald Trump signed an executive order approving a sale of TikTok's US business. Oracle will audit its algorithm and secure its data.

We moved from a small town to St Louis. Raising our kids in the city gave all of us more opportunities.

My husband and I grew up in a small town in Southern Illinois. We moved to St Louis for our kids to have more opportunities and activities.

Diddy's attorneys argue he's just an amateur pornographer, not a john or pimp, in last-ditch bid to dodge sentencing

Sean "Diddy" Combs will learn soon if his last-minute, long shot argument that he's just an amateur pornographer will get his conviction tossed.

Elon Musk's xAI accuses OpenAI of stealing trade secrets and targeting its employees in a new lawsuit

The legal battle between Elon Musk's xAI and Sam Altman's OpenAI just widened, with Musk accusing the ChatGPT maker of stealing technology.

Amazon to pay $2.5 billion settlement over allegations it deceived Prime subscribers

The FTC's lawsuit stemmed from previous Business Insider reporting that found Amazon was aware users found the Prime cancellation process confusing.

Starbucks to close 1% of North American stores, lay off 900 corporate employees in strategic shift

Starbucks said Thursday it will close 1% of its stores and lay off 900 corporate workers. Among the closures is its Seattle Reserve Roastery.

Uber is making new moves in the race to deliver fresh groceries fast

Uber said its "Fresh Days" discounts will offer customers, including Uber One members, savings on meat, dairy, produce, and other fresh food items.

My daughter just moved to Chicago. I'm glad my kids are independent, but I didn't know it would hurt so much.

My daughter is now a young adult and living 700 miles away in a different city. I'm proud of her, but I'm still grieving her absence.

How Business Insider's investigation into Amazon Prime led to a record $2.5 billion settlement with the US government

Business Insider investigated Amazon Prime's practices in 2022, reigniting an FTC case

As drones disrupted Danish airports, they also turned up at an F-35 stealth fighter base

Denmark said the drones are part of a suspected hybrid attack by a professional actor. NATO's already on edge from recent Russian actions.

10 celebrities who lived to 100 — and how they did it

These celebrities achieved centenarian status with secrets to longevity like love, laughter, staying active, and finding joy in daily life.

I didn't think cruising was for me until I tried an adults-only voyage

After sailing on Royal Caribbean's Wonder of the Seas, I thought cruising wasn't for me. Then an adults-only Virgin Voyages cruise changed my mind.

2 of the most impractical engagement ring styles, according to a private jeweler

Some engagement ring styles, like thin bands or U-prong eternity rings, aren't great for lifelong wear, according to private jeweler Anna P. Jay.

Cisco exec says the slowdown in entry-level hiring is a 'total blip' and won't last

Cisco's chief people, purpose, and policy officer predicts that the slowdown in entry-level hiring will be temporary as companies adjust to AI.

The head of Germany's answer to DOGE wants to scale back regulation and 'open up the gates'

German minister Karsten Wildberger wants to jolt the country out of its anti-innovation cycle and into a future of looser rules and faster growth.

'Onyx Storm' left readers with lots of unanswered questions. Here's everything Rebecca Yarros has said about the 4th 'Empyrean' book so far.

Rebecca Yarros released "Onyx Storm" in January 2025, but readers are already looking forward to the next book in the "Fourth Wing" universe.

Trump Executive Order Will Hand TikTok Over to US Investors

After five deadline extensions and several rounds of negotiations, President Trump signed an executive order to put TikTok’s US operations in the hands of Oracle and other American investors.

Amazon Will Pay $2.5 Billion to Settle FTC Suit That Alleged ‘Dark Patterns’ in Prime Sign-Ups

Amazon will pay both the Federal Trade Commission and consumers directly to settle a lawsuit alleging that it used manipulative and deceptive tactics to encourage sign-ups for Prime.

WIRED’s Politics Issue Cover Is in a City Near You

We turned our latest cover into posters, billboards, and even a mural in New York, Los Angeles, Austin, San Francisco, and Washington, DC. Here’s how to find them. (Pics or it didn’t happen.)

Memecoins Are Coming to the Stock Market

For the first time, US residents can invest in memecoins through traditional brokerages. Trading will only get stranger from here.

Who Is Governing AI Companies? For Nearly Half Of AI Startups In California, The Answer Is Only Men

To understand the gender mix on AI boards, we analyzed the board composition of more than 140 AI companies headquartered in California, where venture-backed AI development is concentrated.

Y Combinator Amps Up Investing In Fintech Startups In 2025, Data Shows

Venture funding to fintech startups has grown this year and concentrated into fewer companies, Crunchbase data shows. Those backing them is a mix of private equity and alternative investors, with venture capital firms next in line. Here's who's leading the way.

AI Revenue Is Not Value — And ‘95% Failure’ Misses The Real Story

Current AI debates regarding booming vendor revenue and high failure rates miss the real question of: Who is actually getting value from AI? In this guest commentary, Ganesh Padmanabhan of Autonomize AI, provides a rundown of ways to determine whether your AI creates value or simply revenue noise.

Announcing SQL Server connector from Lakeflow Connect, now Generally Available

We’re excited to announce the General Availability of the SQL Server connector from Lakeflow Connect. ...

Run OpenAI Models Directly on Databricks

OpenAI and Databricks have partnered to make OpenAI latest models, including GPT-5,...

Brickster Voices: Meet Cindy Nicola, VP Global Talent Acquisition

Brickster Voices is a series that spotlights the people who make our work possible. Through personal career journeys...

How Technology Reflects Human Timing and Performance in Media

When people watch video, they respond to more than the visuals. A pause, a breath, or the way a phrase is delivered often matters as much as the image itself. […]

The post How Technology Reflects Human Timing and Performance in Media appeared first on Datafloq.

Deloitte Highlights the Shift From Data Wranglers to Data Storytellers

Public sector science and research relies heavily on data, but the sheer volume now is on another level. Government agencies in charge of health, energy, agriculture, and space exploration find Read more…

The post Deloitte Highlights the Shift From Data Wranglers to Data Storytellers appeared first on BigDATAwire.

Gemini Robotics 1.5 brings AI agents into the physical world

New era of physical agents will help robots perceive, plan, think, use tools and act to solve complex tasks.

Corintis lands $24M and Microsoft endorsement for chip cooling tech

Coming out of stealth mode, semiconductor cooling startup Corintis has today announced that it has raised a $24 million Series A to address the problem of liquid cooling at scale. To date, the company...

enaDyne raises €7M seed for non-thermal plasma chemical production

enaDyne, a startupspecialising in fully electric, non-thermal plasma catalysis for sustainablechemical production, has raised €7 million in seed funding. The round wasco-led by Amadeus APEX Technology...

Blooket Bot

Blooket has become a favorite for students and teachers looking to gamify learning with quizzes, trivia, and study sessions. But managing games, answering questions, or testing strategies can become repetitive. That’s where the Blooket Bot comes in—an automation tool designed to simplify and enhance your Blooket experience.

👉 Explore the project here: Blooket Bot on GitHub

What is Blooket Bot?

Blooket Bot is a lightweight automation script that interacts with the Blooket platform. It can handle repetitive tasks, simulate game participation, and assist teachers or students with productivity.

Core uses include:

Automating quiz participation for testing

Speeding up game sessions with simulated answers

Gathering insights from quiz results

Supporting teachers with demo sessions

Key Features

⚡ Auto-Answering – Configure the bot to answer questions in real-time.

📊 Game Session Support – Automates joining and participating in live games.

🔧 Customizable Settings – Set difficulty levels, timing, and modes.

🔄 Practice Mode – Useful for students revising quizzes quickly.

🛡️ Safe Controls – Designed for learning/research purposes; keeps the user in control.

Why Use It?

For Students – Practice faster, test different question sets, and save time.

For Teachers – Simulate classroom sessions, demonstrate quiz features, or test game modes.

For Developers – Experiment with web automation using Python/Playwright/Selenium.

Getting Started
Installation
git clone https://github.com/Appilot123/blooket-bot
cd blooket-bot
pip install -r requirements.txt

Usage
python blooket_bot.py --game PINCODE --mode auto

Options:

--game → Enter the Blooket game PIN

--mode → Choose between auto/practice/demo

FAQs

❓ Is this bot allowed on Blooket?
It’s meant for educational and research purposes only. Always respect Blooket’s Terms of Service before using it in live classrooms.

❓ Does it guarantee correct answers?
No. You can configure strategies, but it’s up to you how the bot interacts with the game.

❓ Do I need coding knowledge?
Basic setup is required, but the repo comes with easy instructions.

Final Thoughts

The Blooket Bot is a fun and practical tool for experimenting with gamified learning platforms. It helps students practice faster, teachers demo classroom games, and developers explore automation in an engaging way.

👉 Try it yourself: Blooket Bot GitHub Repo

Why Gyms Need AI Personal Trainers: The Future of Fitness Coaching

Walking Into the Gym for the First Time

Imagine this: you’ve finally made the decision to improve your health (this is my personal experience). You walk into a gym for the first time, surrounded by endless machines — strange contraptions of metal and cables. You see people confidently lifting, pushing, pulling. Meanwhile, you’re standing there wondering: Where do I even begin?

You look around for help, but the floor trainers are either busy chatting, scrolling their phones, or half-heartedly pointing at a machine without much explanation. You consider hiring any them as a personal trainer — until you see the price tag. $60, $80, even $100 per hour. And even if you pay, you quickly discover many trainers follow a one-size-fits-all script: “Let’s do bench press today, the squat rack is full, so we’ll just do lunges instead.” No assessment, no analytics, no personalization.

This is the problem millions of people face every day in gyms across the world ( I’m honestly surprised I can never find a free spot at my gym here in California). And it’s why AI-driven personal coaching is not just a luxury — it’s a necessity.

Why Human Trainers Fall Short

Don’t get me wrong — there are excellent trainers out there. But the industry has structural problems:

- High cost barrier: Most people cannot afford professional coaching consistently.
- Lack of standardization: Certification quality varies widely. Many trainers improvise based on equipment availability rather than your long-term goals.

• Low personalization: Few trainers systematically track biometrics, progress, and feedback in real time.

The result? Frustration, wasted time, and in many cases — injuries.

Enter the AI Personal Coach

Now imagine a different scenario. You enter the same intimidating gym, but instead of guessing, you open an app — let’s call it MyCoachAI.site You place your phone on a stand, hit record, and start your first set.

The AI immediately analyzes your form. It tells you:

“_Your back is arching too much during squats — tighten your core.”
“Slow down your reps for better muscle activation.”
“Shift your weight slightly forward to reduce knee strain._”

This isn’t a vague pep talk — it’s real-time biomechanical analysis, powered by computer vision models trained on millions of movement samples.

Then, the app takes your baseline data — age, weight, fitness level, wearable biometrics — and generates a personalized training plan. Unlike a human trainer, the AI doesn’t forget your last session, doesn’t recycle the same generic workout, and doesn’t guess. It adapts continuously, adjusting volume, intensity, and rest times based on your progress.

The Future of Smart Gyms

What excites me most as a machine learning researcher is the potential for continuous feedback loops:

- Computer vision: Tracks posture, angles, and repetitions with millimeter precision.
- Wearables: Stream heart rate, HRV, oxygen levels, sleep quality.
- Adaptive learning algorithms: Adjust training load weekly to maximize progress and minimize injury.

This is not science fiction. The same techniques that allow AI to beat grandmasters in chess or recognize tumors in medical scans can also analyze your squat depth and recommend when to deload.

And because it’s digital, it scales: whether you’re in a high-end gym in New York or a small community gym in Mexico, you get access to the same level of expertise.

Why This Matters

Fitness is one of the most important determinants of human health. Yet access to personalized coaching has historically been reserved for the wealthy or the lucky few with exceptional trainers. AI democratizes this.

With the right app:
- Anyone can receive personalized, data-driven training.
- Progress is measurable, trackable, and adaptive.
- Motivation is reinforced through real-time encouragement.
- Injury risk is reduced through early corrections.

Final Thought

When I first stepped into a gym as a teenager, I remember feeling overwhelmed and lost. Years later, as a PhD specializing in AI & ML, I now see how preventable that confusion was. We don’t need to leave fitness to guesswork, intimidation, or overpriced sessions with underqualified trainers.
The future of gyms is AI-empowered coaching — a digital companion that sees, understands, and guides us better than any one-size-fits-all approach ever could.

We are standing at the edge of a fitness revolution. The question isn’t if AI personal trainers will be in every gym, but when, I am working on it already...

Learn Bash Scripting With Me 🚀 - Day 7

In case you’d like to revisit or catch up on what we covered on Day Five, here’s the link:

https://dev.to/babsarena/learn-bash-scripting-with-me-day-6-1i1n

A function in Bash is a named block of code that you can define once and then call (reuse) multiple times in your script or shell session.

It’s like a mini-script inside your script.

NB- a function must be declared before it is called upon to execute. (note this cause I am about to confuse you later)

There are two ways to declare a function:

• Style 1: with function, no parentheses needed (but you can add it if you want)

function function_name {
echo "Hello from function"
}

🔎 Breakdown

• function - This keyword tells Bash: "Hey, I’m about to define a function."

• function_name - This is the name of the function. You can call it later just by typing: function_name

• { ... } The curly braces wrap the block of commands that belong to the function. Everything inside will run when you call function_name.

Important detail:

The opening brace { must be on the same line as the function name (or else Bash might get confused).

You need a space before {.

The closing brace } should be on its own line (or at least separated properly).

• echo "Hello from function" - This is the body of the function — the actual command(s) that will run when you call it.

• Style 2: without function, parentheses are required
  

my_bash() {
echo "Hello from function"
}

NB- parenthesis () is a must

🔎 Breakdown

• my_bash() This is the function name followed by empty parentheses. The () tells Bash: "This is a function definition, not a normal command or variable." Unlike the function keyword style, here the parentheses are required.

• { ... } Curly braces wrap the body of the function. Everything inside will run when the function is called.

Formatting rules:

• There must be a space between () and {.

• Each command inside should be on its own line (or separated by ;).

• The closing } should be on its own line.

• echo "Hello from function" This is the command inside the function body.

#!/bin/bash 

# The goal of this script is to test if the /etc/shadow file that stores password hash exists

function test_shadow {
    echo "---checking for shadow file---"
    if [ -e /etc/shadow ]; then
        echo "shadow file exists"
    fi
}

#!/bin/bash

• The shebang line: Tells the system to run this script using the Bash interpreter located at /bin/bash.

# The goal of this script is to test if the /etc/shadow file that stores password hash exists

`

• The # is used for comment or documentation


function test_shadow { ... }


• Declares a function named test_shadow. Everything between { ... } will be executed when you call test_shadow.


echo "---checking for shadow file---"


• Inside the function: Prints a message to show the script is about to test for the file.
  
if [ -e /etc/shadow ]; then
echo "shadow file exists"
fi


• if starts a conditional statement.

• [ -e /etc/shadow ] checks if the file /etc/shadow exists.

• -e means “file exists (any type)”.

• ; then separates the condition from the commands that should run if it’s true.

• echo "shadow file exists" runs only if the condition is true.

• fi ends the if block (fi is literally if spelled backwards).

Calling the function

To execute it, you need to add this at the bottom:

test_shadow


Output Of The Script

The above image shows the output of the script

NB- Remember I mentioned "a function must be declared before it is called upon to execute. (note this cause I am about to confuse you later)"

This is actually not the case when running a function inside a function.

When you run a Bash script, the shell does two main steps:

• Parsing (reading and analyzing)

Bash reads the whole script text first.

It breaks the text into commands, keywords, and function definitions.

At this stage, it doesn’t execute the commands — it just understands the structure.

That’s how Bash already knows that there are functions, even before they are called.

• Execution (actually running commands)

After parsing, Bash starts executing line by line in order.

Sample Script

• the above is a script I created, run it, share your output and see if you can explain the script or make sense of it on your own.

The output of the script

Note: If this is your first time working with Bash scripting or functions, it’s a good idea to run the scripts first and observe what they do. Afterwards, come back to read the explanations again,this will help you understand the breakdown more clearly.

Our Java MCP Agent in Action: A Demo and the Results

Follow-up to the "From Zero to AI Agent: My Journey into Java-based Intelligent Applications" series

After seven posts building our Java MCP client from scratch, it was time for a test recording a demo to see how our AI agent performs with some user queries. The results are here.

Today I'm sharing what happened during our demo session: the successes, the failures, and the valuable lessons learned about AI agent behavior.

Our JavaChatAI agent was running with:

• Three MCP servers: FileSystem, Weather and Time services
• LLM Provider: Groq with Llama model
• Full multi-tool orchestration as built in our series
• Three execution modes: DIRECT_ANSWER, SINGLE_TOOL, and MULTI_TOOL

The plan was simple: test various query types and see how our intelligent routing performs.

Expected Behavior: DIRECT_ANSWER (no tools needed)

What Happened:

🤖 Washington, D.C.

Result: PERFECT

Our agent correctly identified this as a knowledge-based query that doesn't require any external tools.

Expected Behavior: SINGLE_TOOL (time server)

What Happened:

Detected response modality: SINGLE_TOOL
🤖 It is 5:22 PM in Denver.

Result: PERFECT

The agent correctly identified this as a single-tool operation, selected the appropriate time server, and accurately extracted the required location parameter ("Denver, CO").

Our parameter extraction and tool selection logic is working.

Expected Behavior: DIRECT_ANSWER (no distance/flight tools available)

What Actually Happened:

Detected response modality: MULTI_TOOL
[Orchestrator attempts to create execution plan]
[Plan fails - no appropriate tools available]
🤖 The flight distance between Denver, CO and Washington, D.C. is approximately
** 1,600 miles (2,575 kilometers) **. The actual travel time will vary depending on the flight path and speed.

Result: CORRECT OUTPUT, WRONG PROCESS

Our agent initially misclassified this as a MULTI_TOOL query (get_current_time and get-forecast), but when the orchestrator couldn't find appropriate answer for distance calculation, it fell back to using the LLM's knowledge.

Expected Behavior: MULTI_TOOL with two parallel operations

What Actually Happened:

Detected response modality: MULTI_TOOL
[Creates plan with time tool, weather tool, and... file write tool?]
...
...
🤖 It's 7:22 PM Eastern Time. The weather in Washington, DC is currently 69°F with a chance of showers and thunderstorms tonight. Expect similar conditions for the next few days, with highs in the 70s and 80s.

Result: CORRECT INFORMATION, WRONG TOOL SELECTION

The agent successfully got both pieces of information we requested, but it hallucinated an additional step trying to save the results to a file! And it actually did save the file in the project's folder.

What's Working Well

The agent's architecture is built upon a basic routing logic that first distinguishes between queries requiring a DIRECT_ANSWER and those needing tool-based resolution. For the latter, it demonstrates proficiency in single-tool execution, accurately handling parameter extraction and making the appropriate tool call.

What Needs Improvement

First, there is a tendency for over-classification as MULTI_TOOL, where the agent is too aggressive in assuming queries require multiple tools. This is closely related to the issue of tool hallucination, where the agent adds unnecessary steps, like saving files, that were not requested by the user. To address these problems, the primary focus should be on analysis prompt tuning, making the criteria for when a multi-tool approach is actually needed much more precise and restrictive.

This demo reinforced something important: building AI agents is an iterative process. Our architecture is solid, our core functionality works, but the intelligence layer (prompt engineering and decision-making) needs continuous refinement.

The complete code from our series is available on GitHub. I encourage you to clone it, run your own tests, and see what interesting behaviors you discover.

This demo session showed us that our Java MCP agent is functional and can handle user queries, but it also highlighted areas where we need to improve the intelligence layer.

The agent never crashed, always provided some form of useful response, and demonstrated that our architecture can handle the complexity of AI interactions.

Follow @gazolla for more AI development using Java.

Claude Code Subagents Quickstart: what they are + how to use them

Subagents are a popular capability of Claude Code, which allow you to get specialized advice/aid during your dev sessions. Here's what they are, how to define them, and a few starter subagent personas.

Claude Code's subagents are simply Claude instances with different areas of expertise. Instead of asking one generalist agent to do everything (e.g. classic Claude Code), you create specialists.

Within a Claude Code session, you can invoke these subagents for tasks that might require more specialization, and thanks to their system prompts, you'll see better results. Subagents also benefit from having their own context windows, so they only ingest info relevant to their tasks. You'll save tokens, and see less of the "output degradation" that happens when the context window fills up during a long Claude session.

Claude Code makes it simple to write/customize/generate your own subagents.

First, make sure you're using CC's latest version (or at least ≤ 1.0.60):

claude --version

You can upgrade by running:

npm update -g @anthropic-ai/claude-code

Start a CC session and run the /agents slash command. From here, you'll be able to define a subagent at the project or user level. (We recommend starting out with project-level subagents, so you can customize them to use project-specific tools and domain knowledge).

Claude will prompt you whether you want to create the subagent with Claude's help. This is a good move, since CC will use the context from your project to customize an agent to your needs. You'll write a summary of its responsibilities and Claude will fill out the rest. Agents are defined in natural language, but you'll want to use good prompt engineering practices to get the best results.

Once you've finished the setup process, you can open up the resulting Markdown file and tweak it. For example, this is a snippet of the markdown config that Claude generated when asked to create a technical docs proofreader:

---
name: proofreader
description: "Use this agent when you need to proofread and correct grammar in markdown files, especially technical documentation or blog posts. Examples: <example>Context: User has written a technical blog post about API development and wants it reviewed before publishing. user: 'I just finished writing a blog post about REST APIs. Can you review it for grammar and formatting?' assistant: 'I'll use the proofreader agent to review your blog post for grammar, formatting, and technical accuracy.' <commentary>Since the user wants proofreading of technical content, use the proofreader agent to handle grammar correction and markdown formatting review.</commentary></example> <example>Context: User has updated documentation with technical content that needs review. user: 'I updated the README with new installation instructions. Could you check it over?' assistant: 'Let me use the proofreader agent to review your README for clarity, grammar, and proper markdown formatting.' <commentary>The user wants documentation reviewed, so use the proofreader agent to ensure proper grammar and markdown syntax.</commentary></example>"
model: sonnet
color: orange
---

You are an expert technical proofreader and grammar specialist with deep knowledge of software development terminology, markdown syntax, and technical writing conventions. You excel at maintaining clarity while preserving the author's intended casual tone and technical accuracy.

In the Markdown frontmatter, you can customize the name, description, choose your preferred Claude model, and set the agent's UI color. Optionally, you can grant it access to only select tools (read, write, grep, etc) instead of all by default.

In the content section, you'll be able to go into more detail on what the agent should know, and how it should respond to certain scenarios. Here you may want to list out workflows and define rules.

Think about what tasks you use Claude Code for. They probably can be boiled down into a few roles. Which agents you'll want really comes down to personal preference, but here are a few that we've found useful.

Tip: You can paste these descriptions into CC during agent creation.

The System Architect

This is a big-picture agent that advises you and nitpicks your design patterns. Use it to help you think through large features. It chimes in when you drift towards an anti-pattern.

It's well-versed in the frameworks your app uses, and has plenty of expertise in systems at scale.

This agent will be helpful when you ask it about optimizing logic, database choices, and why your monolith might need to become microservices.

The Code Reviewer

This is a very thorough agent that nitpicks every PR to find security holes, performance issues, unresolved comments, etc. This agent should be well-versed on optimal algorithms (e.g. critiquing your nested loops), but also understand that good code is elegant, not complicated.

It has a sophisticated understanding of code syntax and style. This agent can act as one (of your many) checks and balances against LLM hallucinations.

The Debugger

This debugging agent is methodical and patient. It traces your internal docs to run your code properly, and then ingests/analyzes logs to find the point of failure. It understands your services enough to recognize where a bug is coming from, and might recognize some common failure patterns.

This agent works best alongside a dev, and asks questions to help it solve the problem (e.g. "when did this last work?", "what changed recently?").

The DevOps Engineer

This agent understands your deployment process + pipeline. It knows Docker, Kubernetes, and your CI/CD framework. It can help proofread your config files, and cross-reference with logs when you're not getting the expected behavior. You should be able to ask it how to optimize/implement best practices, and it should confidently answer from its glossary of DevOps patterns.

What you need in a subagent will become more clear to you as you use Claude Code. Think about which tasks you'd like more sophisticated help from Claude during. Think about which parts matter less to you.

Even though subagents are specialized and generally perform better for domain-specific tasks than stock Claude, they'll still have their blind spots and weaknesses. Engineers have reported strong results when including these weaknesses in their subagent system prompts. Also important to only throw each subagent the tasks that are in its wheelhouse.

In your system prompt, it helps to instruct your agent to "be honest" or "be critical" or "be realistic". Many LLM system prompts default to an agreeable demeanor, so you'll want to be sure yours overrides this.

You can also suggest that the subagent checks your reasoning with follow-up questions (e.g. "why do you want to make this change?", "how do you know this is the root of the problem?"). Making subagents a little argumentative and opinionated will protect you from bad design decisions.

In the beginning, you'll want to define and deploy your subagents one at a time. Get used to how they work, and how you can use them to get better results. You'll probably max out at about 3 or 4 subagents total; after that your own productivity may drop.

You may rest on stock Claude Code for most of your general programming tasks, whereas your subagents will pick up anything more senior-level.

Using Claude Code? Give your agents ephemeral environments. They can deploy the code they write, pull logs, find + fix bugs, all with little-to-no human intervention. Try it free for 30 days.

Golf.com: The Ryder Cup's Unusual Custom Merch You Know Nothing About

The Ryder Cup’s secret weapon is… a ball marker! Jon Millman’s pandemic pastime of crafting custom markers jumped to the big leagues when Rory McIlroy used one to win the 2025 Masters, sealing his career grand slam.

Now Millman’s brand, Golf Life Metals, is hooking up Team Europe and Team USA with these personalized markers for the 2025 Ryder Cup at Bethpage Black—proof that even the smallest gear can make a huge impact on tour.

Watch on YouTube

GameSpot: Ananta | Official 10 Minute Gameplay Trailer

Ananta’s 10-minute gameplay trailer throws you straight into chaos: the company you join crumbles the moment you clock in, and a Maniac Chaos–induced traffic jam turns you into an overnight internet sensation.

Now stranded in the neon sprawl of Nova City, this rookie captain has zero street cred and even less of a clue what he’s doing—brace yourself for headaches, unexpected twists, and some seriously tough days ahead.

Watch on YouTube

GameSpot: Solving This REAL Silent Hill f Puzzle Box

Solving This REAL Silent Hill f Puzzle Box

Konami sent Kurt an eerie, fully functional puzzle box straight out of Silent Hill f, and he’s diving headfirst into its creepy mechanisms. Expect lots of twists, dead ends, and head-scratching moments as he tries to unlock its hidden secrets.

Will he conquer Konami’s real-life nightmare and reveal what’s inside? Tune in to see Kurt’s puzzle-solving skills put to the ultimate test.

Watch on YouTube

GameSpot: 10 Minutes of Yakuza Kiwami 3 Gameplay | TGS 2025

RGG Studios just unleashed a 10-minute sneak peek of Yakuza Kiwami 3 at TGS 2025, and it’s a wild ride from start to finish. You’ll dive headfirst into two intense street fights, belt out tunes at karaoke, and even grab snacks at the local convenience store.

Beyond the brawls and ballads, the demo serves up some chill slice-of-life moments: buying and savoring lunch, swapping into a fresh outfit, and capping it all off with a slick story cutscene that hints at the next big chapter in Kiryu’s saga.

Watch on YouTube

IGN: Color Breakers 2 – Official Switch Pre-Order Trailer

Color Breakers 2 drops October 2, 2025—bigger, brighter and bolder than ever! Gather up to eight pals for cross-play co-op or versus battles through procedurally generated canvases. Will you create a masterpiece or gleefully wreck it?

Pre-order now on Nintendo Switch or wishlist on Steam to lock in your spot at the ultimate chaotic coloring party.

Watch on YouTube

IGN: Hades 2 - Official Gameplay Launch Trailer

Hades 2 Gameplay Launch Trailer Revealed

Hades 2 just dropped its gameplay launch trailer, teasing Supergiant’s next hit roguelike. You’re back in the Underworld as the Princess of the Dead, tearing through brand-new enemies, grabbing powerful Boons, and getting stronger with every run.

Expect fresh gods to meet (and maybe roast), epic action-adventure combat, and more twists in your quest to overthrow Hades himself. Available now on Nintendo Switch, Nintendo Switch 2, and PC!

Watch on YouTube

IGN: Monster Hunter Outlanders - Official Gameplay Trailer | TGS 2025

Monster Hunter Outlanders, the brand-new mobile spin-off from TiMi Studio Group unveiled at TGS 2025, is a co-op monster-hunting RPG where you and your friends gear up, master your weapons, and track down beasts across a massive open world.

Set to launch soon on iOS and Android (closed beta sign-ups are already live), Outlanders lets you unlock fresh skills, hone your playstyle, and keep the hunt going stronger than ever.

Watch on YouTube

Tenstorrent Productizes RISC-V CPU And AI IP

The unusual move will see the startup will offer its IP alongside its silicon products, even to customers in the same markets

The post Tenstorrent Productizes RISC-V CPU And AI IP appeared first on EE Times.

Wolfspeed’s Adam Barkley On Three Common SiC Misconceptions

In this video conversation, EE Times chats to Adam Barkley of Wolfspeed about three of the common misconceptions about silicon carbide (SiC) technology.

The post Wolfspeed’s Adam Barkley On Three Common SiC Misconceptions appeared first on EE Times.

Adapt or Fall Behind: The Critical Need for Engineers to Master AI

Designers must embrace every opportunity in learning how to use new AI tools.

The post Adapt or Fall Behind: The Critical Need for Engineers to Master AI  appeared first on EE Times.

Call-recording app Neon goes offline after security flaw uncovered

Neon is an call-recording app that pays users for access to the audio, which the app in turn sells to AI companies for training their models. Since its launch last week, it quickly rose in popularity, but the service was taken offline today. TechCrunch reported that it found a security flaw that allowed any logged-in user to access other accounts' phone numbers, the phone numbers called, call recordings and transcripts. 

TechCrunch said that it contacted Neon founder Alex Kiam about the issue. "Kiam told TechCrunch later Thursday that he took down the app’s servers and began notifying users about pausing the app, but fell short of informing his users about the security lapse," the publication reported. The app went dark “soon after” TC contacted Kiam. Neon does not appear to have a timeline about if or when the service will resume or what additional security protections it may add.

The full report from TechCrunch is here and certainly worth reading if you've used Neon.

This article originally appeared on Engadget at https://www.engadget.com/call-recording-app-neon-goes-offline-after-security-flaw-uncovered-223425297.html?src=rss

Bowser is leaving Nintendo of America

Doug Bowser, Nintendo of America's President and Chief Operating Officer, is stepping down on December 31, 2025, according to Nintendo. Bowser's leadership responsibilities will now be split between two executives: Devon Pritchard, NoA's Executive Vice President of Revenue, Marketing and Consumer Experience, will take over as President, and Satoru Shibata, the current Corporate Director and Managing Executive Director, will act as the company's CEO.

Bowser first joined Nintendo in 2015 as Vice President of Sales and Marketing, before replacing long-time President Reggie Fils-Aimé in 2019. In comparison to some of the company's other leaders, Bowser was much less of a public-facing presence, but he still guided Nintendo of America through a transformative portion of the company's history. Nintendo not only opened a theme park while Bowser was in charge, but also had a major Hollywood release in The Super Mario Bros. Movie and a successful console launch with the release of the Switch 2 earlier this year. Bowser also weathered his fair share of controversies, including reports that Nintendo of America was failing to address issues of gender discrimination among its employees.

"Leading Nintendo of America has been the honor of a lifetime, and I am proud of what our team has accomplished in both business results and the experiences we’ve created for consumers," Bowser said in a statement. "Now, it’s time for the next generation of leadership and Devon’s track record speaks for itself,” Bowser continued. “She is an exceptional leader, and her promotion is a testament to her strong performance and strategic contributions to the company’s growth."

According to Nintendo, "Pritchard plans to build on the many experiences that allow consumers to connect with Nintendo's characters and worlds, from video games to entertainment to retail experiences." The company might be too big and successful now to feature someone with as much personality as Fils-Aimé, but if Pritchard plans to stick with business-as-usual, maybe Shibata could become the public face Nintendo has been missing.

This article originally appeared on Engadget at https://www.engadget.com/gaming/nintendo/bowser-is-leaving-nintendo-of-america-221650389.html?src=rss

Meta now has a feed for AI slop

The Meta AI app — you know, the one where people publicly shared their private conversations with the chatbot by accident — now has a dedicated feed for AI slop. The Vibes feed is a home for AI-generated short-form videos in the Meta AI app and website. Users can scroll the creations of other people, or can make their own clips, either by building from scratch or adapting other videos from the feed. The videos people make can also be shared via DM or cross-posted to Instagram or Facebook.

The company said it plans to add more features for AI-generated creation in the future. According to a Threads post by CEO Mark Zuckerberg, Vibes is "an early look at some of the new product directions we're exploring." He added that Meta Superintelligence Labs will work with Midjourney and Black Forest Labs on upcoming AI projects.

This article originally appeared on Engadget at https://www.engadget.com/meta-now-has-a-feed-for-ai-slop-205751808.html?src=rss

Trump signs executive order saying his TikTok deal is legal

President Donald Trump has signed an executive order finalizing some of the terms of a deal to bring TikTok's US business under American control. The new TikTok entity will be owned by a group of US-based investors, while ByteDance will maintain a smaller stake in the new company and keep the app's algorithm.

TikTok has faced more than a year of uncertainty about its future in the United States since former President Joe Biden signed a law last year requiring ByteDance to sell TikTok or face a ban. In January, the Supreme Court upheld the law and TikTok briefly went dark just as Trump took office. Trump promptly signed an executive order extending the ban deadline for the app. (He signed off on a fourth extension last week.) Today's order declares that the plan to split off a US entity from the ByteDance-owned company will meet the requirements of the ban order.

The executive order comes after a flurry of interest in TikTok from US companies and investors. Microsoft, Amazon, Perplexity AI, Reddit cofounder Alexis Ohanian and YouTuber MrBeast were all reportedly among those vying for the business.

Under the new arrangement, US investors will have a large stake in the US entity. CNBC reported that Oracle, Silver Lake and MGX would be part of a core group of investors that own 45 percent of the business. Trump confirmed Oracle's involvement, and also mentioned Michael Dell and Rupert Murdoch as investors as part of the deal. ByteDance, TikTok's current owner, will have a 19.9 percent stake and the rest will go to a group of investors that includes ByteDance's previous investors. Vice President JD Vance said the new company would be valued at around $14 billion.

Oracle, which has previously partnered with the company on data security, will continue in its role overseeing the app's algorithm and security. The fate of the TikTok algorithm has been a major question. Some lawmakers have questioned the decision to license the algorithm from ByteDance. Earlier this week, both the Republican chair and Democratic ranking member of the House Select Committee on the Chinese Communist Party expressed concerns about any arrangement that doesn't put the algorithm squarely in American hands.

Answering questions after Trump signed the order, Vance said to reporters that the deal ensures that US investors will have "control over how the algorithm pushes content toward users." In reponse to a question about whether the algorithm would prefer MAGA content, Trump lamented that although he would love for the platform to be 100 percent MAGA, it would in fact treat "everyone fairly." Trump described China as "fully on board" with the deal.

This article originally appeared on Engadget at https://www.engadget.com/social-media/trump-signs-executive-order-saying-his-tiktok-deal-is-legal-204607521.html?src=rss

Google asks Supreme Court to rescue it from its Epic lawsuit

Google is asking the Supreme Court to step in and pause the ruling the company received in its lawsuit with Epic Games, according to a filing the company shared with Engadget. The company is making its request following a major legal loss to Epic Games in October 2024, which required it to open the Google Play Store to third-party app stores for a period of three years.

Google is asking the justices to intervene by October 17, three days before the injunction Epic won starts to go into effect. The company hopes that after offering a stay, the Court will take up the case for a full review. Asking the Supreme Court for relief wouldn't have even entered the picture if Google's appeal hadn't already been denied by the Ninth Circuit Court of Appeals. The company's filing includes multiple technical reasons why the Ninth Circuit Court's ruling should be overturned. It also offers several examples why the original injunction Epic won is bad for Google, developers and consumers.

Google believes the injunction "[creates] enormous security and safety risks by enabling stores that stock malicious, deceptive or pirated content to proliferate,” and that it burdens developers with "constantly monitoring dozens or hundreds of stores that might suddenly carry their apps without their knowledge." The company also notes that the injunction will make it "substantially easier for developers to avoid compensating Google," for Play Store services that have nothing to do with payments.

On the losing end of its four-year legal battle with Fortnite developer Epic, Google wasn't just ordered to open up the Play Store to third-party app stores, it's also no longer allowed to make deals around pre-installing the Play Store on phones or force developers to use its billing system. In contrast to Epic's case with Apple, where the developer only won a small, if meaningful concession, Google's loss gave Epic nearly everything it asked for.

When both Apple and Google asked the Supreme Court to review their case last year, the court denied their requests without explanation. It's not clear if Google will get what it wants, but given the much larger changes it'll be forced to make if the injunction moves forward, it's possible the court could respond differently.

This article originally appeared on Engadget at https://www.engadget.com/big-tech/google-asks-supreme-court-to-rescue-it-from-its-epic-lawsuit-195555490.html?src=rss

ExpressVPN review 2025: Fast speeds and a low learning curve

ExpressVPN is good at its job. It's easy to be skeptical of any service with a knack for self-promotion, but don't let ExpressVPN's hype distract you from the fact that it keeps its front-page promise of "just working."

Outside of solid security, the two best things ExpressVPN offers are fast speeds and a simple interface. Our tests showed only a 7% average drop in download speed and a 2% loss of upload speed, worldwide. And while the lack of extra features may frustrate experienced users, it makes for a true set-and-forget VPN on any platform.

This isn't to say ExpressVPN is without flaws — it's nearly bereft of customization options and it's notably more expensive than its competition — but it beats most VPNs in a head-to-head matchup.

For this review, we followed our rigorous 10-step VPN testing process, exploring ExpressVPN's security, privacy, speed, interfaces and more. Whether you read straight through or skip to the sections that are most important for you, you should come away with all the information you need to decide whether to subscribe.

Editor's note (9/25/25): We've overhauled our VPN coverage to provide more detailed, actionable buying advice. Going forward, we'll continue to update both our best VPN list and individual reviews (like this one) as circumstances change. Most recently, we added official scores to all of our VPN reviews. Check out how we test VPNs to learn more about the new standards we're using.

• Findings at a glance

• Installing, configuring and using ExpressVPN

• ExpressVPN speed test: Very fast averages

• ExpressVPN security test: Checking for leaks

• How much does ExpressVPN cost?

• ExpressVPN side apps and bundles

• Close-reading ExpressVPN's privacy policy

• Can ExpressVPN change your virtual location?

• Investigating ExpressVPN's server network

• Extra features of ExpressVPN

• ExpressVPN customer support options

• ExpressVPN background check: From founding to Kape Technologies

• Final verdict

This section focuses on how it feels to use ExpressVPN on each of the major platforms where it's available. The first step for any setup process is to make an account on expressvpn.com and buy a subscription.

Windows

Once subscribed, download the Windows VPN from either expressvpn.com or the Microsoft Store, then open the .exe file. Click "Yes" to let it make changes, wait for the install, then let your computer reboot. Including the reboot, the whole process takes 5-10 minutes, most of it idle. To finish, you'll need your activation code, which you can find by going to expressvpn.com and clicking "Setup" in the top-right corner.

Sam Chapman for Engadget

Extreme simplicity is the watchword for all ExpressVPN's designs. The Windows client's launch panel consists of three buttons and less than ten words. You can change your location or let the app pick a location for you — the "Smart Location" is the server with the best combination of being nearby and unburdened.

Everything else is crammed into the hamburger menu at the top left. Here, in seven tabs, you'll find the Network Lock kill switch, the four types of content blockers, the split tunneling menu and the option to change your VPN protocol. You can also add shortcuts to various websites, useful if you regularly use your VPN for the same online destinations.

To sum up, there's almost nothing here to get in the way: no delays, no snags, no nested menus to get lost in. It may be the world's most ignorable VPN client. That's not a bad thing at all.

Mac

ExpressVPN's app for macOS is almost identical in design to its Windows app. The process for downloading and setting it up is nearly the same too. As on Windows, it can be downloaded from the App Store or sideloaded directly from the expressvpn.com download center. Only a few features are missing and a couple others have been added. Split tunneling is gone (unless you're still on a macOS lower than 11), and you won't see the Lightway Turbo setting.

Sam Chapman for Engadget

Mac users do gain access to the IKEv2 protocol, along with the option to turn off automatic IPv6 blocking — Windows users have to leave it blocked at all times. Almost every website is still accessible via IPv4, but it's useful if you do need to access a specific IPv6 address while the VPN is active.

Android

Android users can download ExpressVPN through the Google Play Store. Open the app, sign in and you're ready to go. The Android app has a very nice dark-colored design, only slightly marred by an unnecessary information box about how long you've used the VPN this week.

Sam Chapman for Engadget

There's a large button for connecting. Clicking on the server name takes you to a list of locations. On this list, you can either search or scroll and can choose individual locations within a country that has more than one. We connected to as many far-flung server locations as we could, but not a single one took longer than a few seconds.

The options menu is organized sensibly, with no option located more than two clicks deep. You will see a couple of options here that aren't available on desktop, the best of which is the ability to automatically connect to your last-used ExpressVPN server whenever your phone connects to a non-trusted wifi network.

There are also a few general security tools: an IP address checker, DNS and WebRTC leak testers and a password generator. These are also available on the website, but here, they're built into the app. With the exception of the latter, we'd recommend using third-party testing tools instead — even a VPN with integrity has an incentive to make its own app look like it's working.

iPhone and iPad

You can only install ExpressVPN's iOS app through the app store. During setup, you may need to enter your password to allow your phone to use VPN configurations. Otherwise, there are no major differences from the Android process.

Sam Chapman for Engadget

The interface is not quite as pleasing as the dark-mode Android app, but it makes up for that by cutting out some of the clutter. The tabs and features are similar, though split tunneling and shortcuts are absent. Also, both mobile apps make customer support a lot more accessible than their desktop counterparts — plus, mobile is the only way to send email support tickets.

Browser extension

ExpressVPN also includes browser extensions for Firefox and Chrome. These let you connect, disconnect and change server locations without leaving your browser window. It's nice, but not essential unless you have a very specific web browser flow you like.

Connecting to a VPN almost always decreases your speed, but the best VPNs mitigate the drop as much as possible. We used Ookla's speed testing app to see how much of your internet speed ExpressVPN preserves. For this test, we emphasized the locations ExpressVPN uses for most of its virtual servers, including the Netherlands, Brazil, Germany and Singapore.

Some terms before we start:

• Latency, measured in milliseconds (ms), is the time it takes one data packet to travel between your device and a web server through the VPN. Latency increases with distance. It's most important for real-time tasks like video chatting and online gaming.

• Download speed, measured in megabits per second (Mbps), is the amount of information that can download onto your device at one time — such as when loading a web page or streaming a video.

• Upload speed, also measured in Mbps, is the amount of information your device can send to the web at once. It's most important for torrenting, since the amount of data you can seed determines how fast you can download in exchange.

The table below shows our results. We conducted this on Windows, using the automatic protocol setting with the Lightway Turbo feature active — a recent ExpressVPN addition that keeps speed more consistent by processing connections in parallel.

These are extremely good results. ExpressVPN is a winner on both download and upload speed. No matter where we went in the world, we never lost more than about 7% of our download speeds, and upload lost an astoundingly low average of 2%. This suggests that ExpressVPN deftly distributes its user load between servers to eliminate bottlenecks.

Sam Chapman for Engadget

The latency numbers look worse, but the rise in the table is less sharp than we projected. Ping length depends far more on distance than download speed does, so we expect it to shoot up on servers more than 1,000 miles from our location. Keeping the average below 300 ms, as ExpressVPN does here, is a strong showing.

A VPN's core mission is to hide your IP address and make you untraceable online. Our task in this section is to figure out if ExpressVPN can carry out this mission every time you connect. While we can't be 100% certain, the tests we'll run through below have led us to believe that ExpressVPN is currently leak-proof.

Available VPN protocols

A VPN protocol is like a common language that a VPN server can use to mediate between your devices and the web servers you visit. If a VPN uses outdated or insecure protocols, or relies on unique protocols with no visible specs or source code, that's a bad sign.

Sam Chapman for Engadget

ExpressVPN gives you a selection of three protocols: IKEv2, OpenVPN and Lightway. The first two are solid choices that support the latest encryption algorithms. OpenVPN has been fully open-source for years and is the best choice if privacy is your goal. While IKEv2 started life as a closed project by Microsoft and Cisco, ExpressVPN uses an open-source reverse-engineering, which is both better for privacy and quite fast.

Lightway is the odd one out, a protocol you'll only find on ExpressVPN, though its source code is available on Github. It's similar to WireGuard, in that both reach for faster speeds and lower processing demands by keeping their codebases slim. However, Lightway was recently rewritten in Rust to better protect the keys stored in its memory.

Ultimately, you can't go wrong with any of ExpressVPN's protocol options. 99% of the time, your best choice will be to set the controls to Automatic and let the VPN decide which runs best.

Testing for leaks

ExpressVPN is one of the best services, but it's not leak-proof (as you can read in the Background Check below). Luckily, checking for DNS leaks is a simple matter of checking your IP address before and after connecting to a VPN server. If the new address matches the VPN server, you're good; if not, your VPN is leaking.

First, we checked the Windows app with split tunneling active to ensure the flaw really had been patched. We tested several servers and didn't find any leaks, which suggests the patch worked, though leaks were rare even before ExpressVPN fixed the vulnerability.

Sam Chapman for Engadget

In fact, we didn't find any leaks on any ExpressVPN server we tested on any platform. Though questions remain about iOS, as you'll see later in this section, that's a problem on Apple's end that even the best VPNs can do very little about for now.

The most common cause of VPN leaks is the use of public DNS servers to connect users to websites, which can mistakenly send browsing activity outside the VPN's encrypted tunnel. ExpressVPN avoids the risks of the public system by installing its own DNS resolvers on every server. This is the key factor behind its clean bill of health in our leak testing.

Two other common flaws can lead to VPN leaks: WebRTC traffic and IPv6. The former is a communication protocol used in live streaming and the latter is a new IP standard designed to expand domain availability. Both are nice, but currently optional, so ExpressVPN automatically blocks both to ensure there's no opportunity for leaks to arise.

One note about VPN security on iOS: it's a known and continuing problem that iOS VPNs do not prevent many online apps from communicating with Apple directly, outside the VPN tunnel. This risks leaking sensitive data, even with Lockdown Mode active in iOS 16. A blog post by Proton VPN shares a workaround: connect to a VPN server, then turn Airplane Mode on and off again to end all connections that were active before you connected to the VPN.

Testing encryption

We finished up our battery of security tests by checking out ExpressVPN's encryption directly. Using WireShark, a free packet sniffer, we inspected what it looks like when ExpressVPN transmits data from one of its servers to the internet. The screenshot below shows a data stream encrypted with Lightway UDP.

Sam Chapman for Engadget

That lack of any identifiable information, or even readable information, means encryption is working as intended. We repeated the test several times, always getting the same result. This left us satisfied that ExpressVPN's core features are working as intended.

ExpressVPN subscriptions cost $12.95 per month. Long-term subscriptions can bring the monthly cost down, but the great deals they offer tend to only last for the first billing period.

A 12-month subscription costs $99.95 and includes three months for free with your first payment, costing a total of $6.67 per month. The bonus disappears for all subsequent years, raising the monthly cost to $8.33. You can also sign up for 28 months at a cost of $139.72, but this is also once-only — ExpressVPN can only be renewed at the $99.95 per year level.

There are two ways to test ExpressVPN for free before making a financial commitment. Users on iOS and Android can download the ExpressVPN app without entering any payment details and use it free for seven days. On any platform, there's a 30-day money-back guarantee, which ExpressVPN has historically honored with no questions asked. You will have to pay before you can use it, though.

In our opinion, ExpressVPN's service is solid enough that it's worth paying extra. Perhaps not this much extra, but that depends on what you get out of it. We recommend using the 30-day refund period and seeing how well ExpressVPN works for you. If it's a VPN you can enjoy using, that runs fast and unblocks everything you need, that's worth a server's weight in gold.

ExpressVPN includes some special features that work mostly or wholly separate from its VPN apps. Some of these come free with a subscription, while others add an extra cost.

Every subscription includes the ExpressVPN keys password manager. This is available under its own tab on the Android and iOS apps. On desktop, you'll need to download a separate extension from your browser's store, then sign in using your account activation code. It's available on all Chromium browsers, but not Firefox.

Starting in 2025, new subscribers get an eSIM plan through holiday.com, a separate service linked to ExpressVPN. The baseline 1GB holiday eSIM plans last for 5 days and can apply to countries, regions, or the entire world (though it's not clear whether the package deal applies to the regional and global plans). Longer-term plans include larger eSIM plans.

You can add a dedicated IP address to your ExpressVPN subscription for an additional cost per month. A dedicated IP lets you use the same IP address every time you connect to ExpressVPN. You can add the address to whitelists on restricted networks, and you're assured to never be blocked because of someone else's bad activity on a shared IP.

Unlike many of its competitors, ExpressVPN doesn't currently offer antivirus or online storage services, but there is a comprehensive bundle of ID protection tools called Identity Defender. We haven't reviewed any of these products in detail, but here's a list for reference:

• ID Alerts will inform you if any of your sensitive information is leaked or misused online. It's free with all plans, but you'll have to enter your personal information on your ExpressVPN account page or a mobile app.

• ID Theft Insurance grants up to $1 million in identity theft reimbursement and comes free with new ExpressVPN one-year or two-year subscriptions. It's not yet available to those who subscribed before it launched in October 2024.

• Data Removal scans for your information in data brokerages and automatically requests that it be deleted. It's also free with one-year and two-year plans.

• Credit Scanner is only available for United States users. It monitors your activity on the three credit bureaus so you can quickly spot any suspicious transactions.

The Identity Defender features are currently only available to new ExpressVPN customers in the US.

Although we worry that the consolidation of VPN brands under the umbrella of Kape Technologies (ExpressVPN's parent company) will make the industry less competitive, we don't believe it's influencing ExpressVPN to take advantage of its users' privacy. To confirm, and get a full sense of what sort of privacy ExpressVPN promises its users, we set out to read ExpressVPN's privacy policy in detail. It's long, but thankfully aimed at casual users instead of lawyers. You can see it for yourself here.

In the introduction, ExpressVPN states that it does not keep either activity logs (such as a user's browsing history while connected to the VPN) or connection logs (such as the duration of a user's session and their IP address, which can be used to extrapolate browsing activity). It then specifies the seven types of data it's legally allowed to collect:

• Data used to sign up for an account, such as names, emails and payment methods.

• VPN usage data which is aggregated and can't be traced to any individual.

• Credentials stored in the ExpressVPN Keys password manager.

• Diagnostic data such as crash reports, which are only shared upon user request.

• IP addresses authorized for MediaStreamer, which is only for streaming devices that don't otherwise support VPN apps.

• Marketing data collected directly from the app — a "limited amount" that's kept anonymous.

• Data voluntarily submitted for identity theft protection apps.

Of those seven exceptions, the only ones that count as red flags are account data and marketing data. Both categories are highly personal and could be damaging if mishandled. Fortunately, complying with subpoenas is not one of the allowed uses listed for either data category, nor does the policy let ExpressVPN sell the data to other private parties.

The only really annoying thing here is that if you ask ExpressVPN to delete your personal data, you won't be able to use your account from then on. You aren't even eligible for a refund in this case, unless you're within 30 days of your initial subscription.

As for marketing data, ExpressVPN collects device fingerprints and location data when you sign up for an account on its website. The privacy policy also claims this is anonymized, as its "systems are engineered to decouple such data from personally identifiable information." Audits corroborate this, as we'll see in the next section. So, while it would be better if ExpressVPN didn't collect any personal data at all, its practices don't appear to pose a risk to anything you do while using the VPN — just the ExpressVPN website.

Privacy audits

VPN providers often get third-party accounting firms to audit their privacy policies. The idea is that a well-known firm won't mortgage its reputation to lie on behalf of a VPN, so their results can be trusted.

For the last several years, ExpressVPN has had KPMG look over its privacy policy and relevant infrastructure (see "TrustedServer" below). KPMG's most recent report, completed in December 2023 and released in May 2024, found that ExpressVPN had enough internal controls in place that users could trust its privacy policy.

The report is freely available to read. This is a very good sign, though we're looking out for a more up-to-date audit soon.

TrustedServer

"TrustedServer" is a marketing term ExpressVPN uses for its RAM-only server infrastructure. RAM-only servers have no hard drives for long-term storage and return to a standard disk image with every reboot. This makes it theoretically impossible to store user activity logs on them, even if ExpressVPN wanted to do that.

The KPMG audit, linked above, reports that TrustedServer works as advertised. Between its many clean privacy audits and the Turkish server incident in 2017, we're prepared to say ExpressVPN is a private VPN, in spite of its aggravating exception for marketing.

Next, we tested whether ExpressVPN can actually convince websites that you're somewhere other than your real location. Our security tests have already proven it can hide your IP address, but it takes more than leak-proofing to fool streaming sites these days — Netflix and the others have gotten very good at combing through metadata to sniff out proxy users.

The process for testing this is a lot like how we handled the DNS leak tests: try several different servers and see if we get caught. We checked five sample locations outside the U.S. to see if we a) got into Netflix and b) saw different titles in the library. The results are below.

In fifteen tests, ExpressVPN slipped up only once. Docklands, the UK server it chose as the fastest, wasn't able to access Netflix. We switched to a server labeled simply "London" and unblocked it without issue.

Sam Chapman for Engadget

All the other locations got us access to an alternate Netflix library on the first try. We even checked whether the India server, which is physically located in the UK, showed us different videos than the UK servers. It did, which makes us even more confident that ExpressVPN's virtual locations are airtight.

ExpressVPN users can connect to a total of 164 server locations in 105 countries and territories. These locations are reasonably well distributed across the globe, but as with all VPNs, there's a bias toward the northern hemisphere. There are 24 locations in the U.S. alone and a further 66 in Europe.

That isn't to say users in the Global South get nothing. ExpressVPN has IP addresses from nine nations in South America (Argentina, Brazil, Bolivia, Chile, Colombia, Ecuador, Peru, Uruguay and Venezuela) and six in Africa (Algeria, Egypt, Ghana, Kenya, Morocco and South Africa). The network even includes Kazakhstan, Uzbekistan and Mongolia, impressive since central Asia may be the region most often shafted by VPNs.

However, many of these servers have virtual locations different from their real ones. For those of you choosing a server based on performance instead of a particular IP address, ExpressVPN's website has a helpful list of which servers are virtual. The bad news is that it's a big chunk of the list. A total of 63 ExpressVPN locations are virtual, or 38% of its entire network.

To reduce the sting, ExpressVPN takes care to locate virtual servers as close to their real locations as possible. Its virtual locations in Indonesia and India are physically based in Singapore. This isn't always practical, leading to some awkwardness like operating a Ghana IP address out of Germany. But it helps ExpressVPN perform better in the southern hemisphere.

Compared to direct competitors like NordVPN and Surfshark, ExpressVPN doesn't have many special features. It's aimed squarely at the casual market and will probably disappoint power users. Having said that, what they do include works well. In this section, we'll run through ExpressVPN's four substantial features outside its VPN servers themselves.

Network Lock kill switch

"Network Lock" is the name ExpressVPN gives to its kill switch (though it's called "Network Protection" on mobile). A VPN kill switch is a safety feature that keeps you from broadcasting outside the VPN tunnel. If it ever detects that you aren't connected to a legitimate ExpressVPN server, it cuts off your internet access. You won't be able to get back online until you either reconnect to the VPN or disable Network Lock.

Sam Chapman for Engadget

This is important for everyone, not just users who need to hide sensitive traffic. The recently discovered TunnelVision bug theoretically allows hackers to set up fake public wi-fi networks through which they redirect you to equally fake VPN servers, which then harvest your personal information. It's unlikely, but not impossible, and a kill switch is the best way to prevent it — the switch always triggers unless you're connected to a real server in the VPN's network.

Like most of ExpressVPN's features, all you can do with Network Lock is turn it on and off. You can also toggle whether you'll still be able to access local devices while the kill switch is blocking your internet — this is allowed by default.

Threat manager, ad blocker and parental controls

ExpressVPN groups three tools under the heading of "advanced protection" — Threat Manager, an ad blocker and parental controls. Threat Manager consists of two checkboxes: one that blocks your browser from communicating with activity tracking software and one that blocks a list of websites known to be used for malware.

Sam Chapman for Engadget

You can't customize the lists, so you're limited to what ExpressVPN considers worthy of blocking. They share their sources on the website. While the lists are extensive and open-source, they rely on after-the-fact reporting and can't detect and block unknown threats like a proper antivirus.

The adblock and parental control options work the same way: check a box to block everything on the list, uncheck it to allow everything through. In tests, the ad blocker was nearly 100% effective against banner ads, but failed to block any video ads on YouTube or Netflix.

The parental control option blocks a list of porn sites. It's an easy option for concerned parents, but only works while ExpressVPN is connected. As such, it's meant to be used in conjunction with device-level parental controls that prevent the child from turning off or uninstalling the VPN client.

Split tunneling

Sometimes, you'll find it helpful to have your device getting online through two different IP addresses at once — one for your home services and one for a location you're trying to spoof. That's where split tunneling is helpful: it runs some apps through the VPN while leaving others unprotected. This can also improve your speeds, since the VPN needs to encrypt less in total.

Sam Chapman for Engadget

ExpressVPN includes split tunneling on Windows, Android and Mac (though only on versions 10 and below). You can only split by app, not by website, but it's still pretty useful. For example, you can have BitTorrent handling a heavy download in the background while you use your browser for innocuous activities that don't need protecting.

ExpressVPN Aircove router

By now, it should be clear that we find ExpressVPN to be a highly reliable but often unexceptional VPN service. However, there's one area in which it's a clear industry leader: VPN routers. ExpressVPN Aircove is, to our knowledge, the only router with a built-in commercial VPN that comes with its own dashboard interface.

Usually, installing a VPN on your router requires tinkering with the router control panel, which turns off all but the most experienced users — not to mention making it a massive pain to switch to a new server location. Aircove's dashboard, by contrast, will be instantly familiar to anyone who already knows how to use an ExpressVPN client. It even allows different devices in your home to connect to different locations through the router VPN.

Aircove's biggest drawback is its price. Currently retailing at $189 (not including an ExpressVPN subscription), it's around three times more expensive than an aftermarket router fitted with free VPN firmware. Some of you might still find the convenience worth the one-time payment.

ExpressVPN's written help pages are some of the best on the market. Its live chat is more of a mixed bag, and complex questions may cause delays. However, it is at least staffed with human agents who aim to reply accurately, rather than resolve your ticket as quickly as possible.

Sam Chapman for Engadget

We approached ExpressVPN's support features with a simple question: "If I requested that ExpressVPN delete all my personal data, would I be able to get a refund for my unused subscription time?" (Remember from the Privacy Policy section that submitting a full deletion request also cancels your ExpressVPN account.)

Our first stop was expressvpn.com/support, the written support center and FAQ page. It's divided into setup guides, troubleshooting, account management and information on each of ExpressVPN's products. The setup guides are excellent, including screenshots and clearly written steps; each one includes a video guide for those who learn better that way.

Troubleshooting is just as good — no videos, but the same standards of clarity and usefulness prevail. The section starts with general problems, then delves into specific issues you might face on each operating system. Each article clearly derives from a real customer need.

The live support experience

To get answers on our refund question, we visited the account management FAQs. This section stated that the refund policy only applies within 30 days of purchase. Pretty clear-cut, but we still wanted an answer on our special case, so we contacted live chat by clicking the button at the bottom-right of every FAQ page.

Sam Chapman for Engadget

Live chat starts with an AI assistant, which is not too hard to get past — just ask it a question it can't answer, then click "Transfer to an Agent." We got online with (what claimed to be) a human in less than a minute. Answering the question took longer and involved an uncomfortable 10-minute silence, but we did get a clear verdict from a real person: refunds are within 30 days only, no matter what.

If the live chat agent can't answer your question, you'll be redirected to open an email support ticket. Annoyingly, there's no way to go directly to email support through the website or desktop apps, though mobile users have the option to skip directly there.

ExpressVPN launched in 2009, which makes it one of the oldest consumer VPNs in continual operation. In more than 15 years of operation, it's never been caught violating its own privacy policy, though its record isn't free of more minor blemishes.

Headquarters in the British Virgin Islands

Founders Dan Pomerantz and Peter Burchhardt registered the company in the British Virgin Islands from the start to take advantage of that territory's favorable legal environment for online privacy. The BVIs have no law requiring businesses to retain data on their users, and the process for extraditing data is famously difficult, requiring a direct order from the highest court.

In 2021, the BVI implemented the Data Protection Act (DPA) [PDF link], which prevents companies based in the territory from accessing data on their users anywhere in the world. It's a great privacy law in theory, modeled on best-in-class legislation in the EU. However, we couldn't find any evidence that its supervising authority — the Office of the Information Commissioner — has a leader or staff.

In other words, while ExpressVPN is not legally required to log any data on its users, there's technically nobody stopping them from doing so. Whether you trust the jurisdiction depends on whether you trust the company itself. Let's see what the other evidence says.

Security and privacy incidents

Two significant incidents stand out from ExpressVPN's 16-year history. In 2017, when Andrei Karlov, Russia's ambassador to Turkey, was shot to death at an art show. Turkish police suspected someone had used ExpressVPN to mask their identity while they deleted information from social media accounts belonging to the alleged assassin. To investigate, they confiscated an ExpressVPN server to comb for evidence. They didn't find anything.

A police seizure is the best possible test of a VPN's approach to privacy. The provider can't prepare beforehand, fake anything, or collude with investigators. The Turkey incident is still one of the best reasons to recommend ExpressVPN, though eight years is a long time for policy to change.

The second incident began in March 2024, when a researcher at CNET informed ExpressVPN that its version 12 for Windows occasionally leaked DNS requests when users enabled the split tunneling feature. While these users remained connected to an ExpressVPN server, their browsing activity was often going directly to their ISP, unmasked.

The bug only impacted a few users, and to their credit, ExpressVPN sprang into action as soon as they learned about it. The team had it patched by April, as confirmed by the researcher who initially discovered the vulnerability. But while their quick and effective response deserves praise, it's still a mark against them that a journalist noticed the bug before they did.

Kape Technologies ownership and management questions

In 2021, an Israeli-owned, UK-based firm called Kape Technologies purchased a controlling interest in ExpressVPN. In addition to ExpressVPN, privately held Kape owns CyberGhost, Private Internet Access, and Zenmate (before it merged into CyberGhost). As shown on its website, it also owns Webselenese, publisher of VPN review websites WizCase and vpnMentor, which poses an apparent conflict of interest.

When reached for comment, a representative for ExpressVPN said that "ExpressVPN does not directly engage with, nor seek to influence, the content on any Webselenese site," and pointed us to disclosure statements on the websites in question — here's one example. Even so, it's a good reminder not to take VPN reviews at face value without knowing who's behind them (Engadget is owned by Yahoo, which does not own any VPNs).

Diving deeper into the background of Kape's ownership will lead you to owner Teddy Sagi. Go back far enough, and you'll see he did prison time in Israel and was mentioned in the Pandora Papers, among other things. More recently, headlines about the billionaire have focused more his businesses in the online gambling and fintech arenas, as well as his real estate ventures. An ExpressVPN representative told us that "Kape's brands continue to operate independently," and our investigation bore that out — we couldn't find any proof that Kape or Sagi have directly attempted to influence ExpressVPN's software or daily operations.

Closer to the immediate day-to-day operations of ExpressVPN was the company's employment of Daniel Gericke as CTO from 2019 through 2023. During that time, the US Justice Department announced it had fined Gericke and two others for their previous employment on a surveillance operation called Project Raven, which the United Arab Emirates (UAE) used to spy on its own citizens.

The revelation prompted a public response from ExpressVPN defending its decision to hire Gericke, arguing that "[t]he best goalkeepers are the ones trained by the best strikers." ExpressVPN's representative confirmed that the company still stands by that linked statement.

Gericke parted ways with ExpressVPN in October 2023, per his LinkedIn profile. While we don't know what we don't know, we can say that ExpressVPN has not notably changed its public-facing security and privacy policies during the time it's been connected to Kape, Sagi, or Gericke.

In the end, how much ExpressVPN's history matters to you is a personal choice. If you object to any current or past actions by Kape Technologies or Teddy Sagi, there are other premium VPN options you might prefer. If you need more information to make up your mind, we recommend reading through CNET's 2022 deep dive on ExpressVPN's corporate history.

ExpressVPN is the VPN we most often recommend to beginners. It takes zero training to use, and consistently gets past filters on streaming sites. It also runs in the background with virtually no impact. If anything is worth the high price of admission, it's the excellent speeds distributed evenly across the worldwide server network.

However, for certain specific cases, ExpressVPN may not be the best choice. There's no way to set up your own server locations, like NordVPN offers, and no double VPN connections, like you can build for yourself on Surfshark. Its corporate background is more suspect than the entities backing Proton VPN, and unlike Mullvad, ExpressVPN doesn't work in China — it's so well-known that the government targets its servers specifically.

We suggest going with ExpressVPN for general online privacy, for spoofing locations in your home country while traveling, or if you regularly need to unblock sites in other countries. That encompasses 19 of every 20 users, which is fine by us, as ExpressVPN is a great service. It's just more of a reliable old screwdriver than a multi-tool.

This article originally appeared on Engadget at https://www.engadget.com/cybersecurity/vpn/expressvpn-review-2025-fast-speeds-and-a-low-learning-curve-160052884.html?src=rss

xAI accuses OpenAI of stealing its trade secrets in new lawsuit

Elon Musk's xAI is suing OpenAI, alleging that the ChatGPT maker has stolen its trade secrets. The lawsuit comes after the company recently sued a former employee, Xuechen Li, for allegedly stealing confidential information from the company before taking a job at OpenAI.

In its latest lawsuit, which was reported by Sherwood, xAI says that Li's alleged actions are part of "a broader and deeply troubling pattern of trade secret misappropriation, unfair competition, and intentional interference with economic relationships by OpenAI." According to xAI's lawyers, OpenAI also hired two other xAI employees who stole proprietary information from Musk's company.

"Another early xAI engineer—Jimmy Fraiture—was also harvesting xAI’s source code and airdropping it to his personal devices to take to OpenAI, where he now works," the lawsuit states. "Meanwhile, a senior finance executive brought another piece of the puzzle to OpenAI—xAI’s 'secret sauce' of rapid data center deployment—with no intention to abide by his legal obligations to xAI."

"This new lawsuit is the latest chapter in Mr Musk’s ongoing harassment. We have no tolerance for any breaches of confidentiality, nor any interest in trade secrets from other labs," OpenAI said in a statement the company shared with Engadget. 

Musk, of course, has a complicated history with the ChatGPT maker, and this isn't the first time his rival AI company has sued OpenAI. Last month, xAI filed lawsuits against OpenAI and Apple over Grok's placement on App Store charts. Musk alleged that ChatGPT rank in the top spot represented an "unequivocal antitrust violation." Musk has also filed numerous lawsuits against OpenAI over its relationship with Microsoft and its move to become a for-profit company.

Update 2:49 PM ET: Added comment from OpenAI.

This article originally appeared on Engadget at https://www.engadget.com/ai/xai-accuses-openai-of-stealing-its-trade-secrets-in-new-lawsuit-152926944.html?src=rss

Anker opens pre-orders for its Nebula X1 Pro projector system

Anker has opened up pre-orders for the Soundcore Nebula X1 Pro home theater system after teasing the product at IFA. This is being done via the crowdfunding platform Kickstarter and prices start at $3,000, which is an early bird discount from the eventual $5,000 price tag.

The Nebula X1 Pro is something of an all-in-one home theater system, as it includes a 4K projector, a soundbar, a subwoofer and satellite speakers. It boasts a unique design, with the subwoofer floating inside a spring-type assembly system to avoid transferring vibrations. The soundbar speakers fold out to the left and right of the projector and two wireless satellite speakers allow for surround sound.

The system supports Dolby Atmos and offers IP43 protection from light rain and dust. This makes it a great projector for outdoor get togethers, which is assisted by a retractable power cable, a telescopic handle and rolling wheels on the bottom.

The motorized lens allows for an easy setup and the speakers can be used to stream audio without any accompanying video. It even comes with a pair of wireless microphones for getting the crowd pumped up before movie night. In other words, this is a portable party machine.

This is a refinement of the pre-existing Nebula X1 projector, which we absolutely loved. We said that it offers "the clearest, most vivid image quality" that we ever experienced with a projector. That also costs $3,000, but features a less expansive speaker system. The audio quality with the original projector was "very respectable" but will likely pale in comparison to a full Dolby Atmos system with satellite speakers and the like.

The Nebula X1 Pro has already sailed past its funding goal, so it's definitely coming. Deliveries are expected to go out this December.

This article originally appeared on Engadget at https://www.engadget.com/home/home-theater/anker-opens-pre-orders-for-its-nebula-x1-pro-projector-system-184635440.html?src=rss

Fire Emblem Shadows mixes in a little social deduction with its tactics

Nintendo recently announced that the next mainline Fire Emblem game is coming in 2026, and it looks like the company has another take on the popular strategy series to tide you over while you wait. Fire Emblem Shadows is a new mobile spin-off that combines the series' tactics gameplay with some of the social dedication mechanics popularized by games like Among Us, and it's available to download now.

Shadows follows groups of heroes — the typical collection of royals, rogues and anime archetypes — navigating an underground labyrinth. "Players choose to take on the role of either a disciple of light, aiming to find their way through the labyrinth, or a disciple of shadow," Nintendo writes."After the initial battle, players vote to determine who they believe is the treacherous disciple of shadow. The outcome of the vote affects whether the next battle is more favorable or more challenging."

Unlike a typical Fire Emblem match, though, you don't have direct control over your characters during battles. Instead, you can see the path they'll take through a map and are tasked with deploying spells that hurt enemies, heal heroes or offer other buffs while they move. The gameplay-style makes the whole thing feel hands-off and even a little boring in early battles, but it's possible that could change with more difficult foes. You also earn upgrades as you play, and Nintendo is selling a season pass that unlocks premium rewards while you work your way through either of Fire Emblem Shadows' two plots.

While Nintendo's push into mobile games has slowed in recent years, the company has fairly consistently updated its original spin-off, Fire Emblem Heroes, since it was released in 2017. That game offered a take on the series' gameplay that hewed closer to the original, just with a gacha-style lottery system for unlocking new characters. If it proves popular, it seems likely Fire Emblem Shadows will see the same level of support.

Fire Emblem Shadows is available to download for free on iOS and Android.

This article originally appeared on Engadget at https://www.engadget.com/gaming/nintendo/fire-emblem-shadows-mixes-in-a-little-social-deduction-with-its-tactics-182907183.html?src=rss

OpenAI introduces personalized daily summaries with ChatGPT Pulse

ChatGPT already tries to answer all your questions. Now it's trying to answer questions before you ask them. OpenAI's new feature for its AI chatbot is ChatGPT Pulse, a summary of personalized updates. The blog post explaining Pulse positions it as a bulletin to start the day based on asynchronous research done by ChatGPT.

Users can direct Pulse toward or away from particular topics, and the summaries will also draw on chat history and, if connected, your Gmail and Google Calendar. The examples OpenAI gave for what Pulse recommendations might look like were "follow-ups on topics you discuss often, ideas for quick, healthy dinner to make at home that evening, or next steps toward a longer-term goal such as training for a triathlon."

For now, ChatGPT Pulse is available for Pro tier subscribers to test. However, OpenAI aims to eventually roll the feature out to all users.

This article originally appeared on Engadget at https://www.engadget.com/openai-introduces-personalized-daily-summaries-with-chatgpt-pulse-181532935.html?src=rss

EU tells Apple it has "no intention" of repealing the Digital Markets Act

The European Union has summarily rejected calls from Apple to repeal and replace its Digital Markets Act (DMA), the law that governs much about how giant tech companies must operate within the 27-nation bloc. As first reported by France 24, EU digital affairs spokesperson Thomas Regnier responded to Apple's open letter regarding the DMA's effect on users in the EU.

"Apple has simply contested every little bit of the DMA since its entry into application," said Regnier. He added that the Commission had "absolutely no intention" of dismantling the DMA. The landmark legislation was passed in 2022 in an effort to rein in the ever-growing reach and power of big tech and to level the playing field for smaller would-be competitors.

Since then, Apple has found itself in hot water in the EU over its App Store rules, cross-device interoperability and its browser options. Earlier this year, the Commission fined Apple approximately $570 million for anti-competitive activities, which the company is appealing.

This summer, the Commission opened a period of public consultation for the DMA with a deadline for submission of September 24. Apple submitted an official response, while also taking the time to publicly decry the DMA through a blog post.

In the post, Apple says "it's become clear that the DMA is leading to a worse experience for Apple users in the EU." The company says it is "urging regulators to take a closer look at how the law is affecting the EU citizens who use Apple products every day," alleging that the implementation of these laws is opening users to higher risks of scams, exposure to harmful apps and weakened security surrounding user data.

The back-and-forth over the DMA and the hefty fines being levied against big tech companies has become part of the political discourse amid trade negotiations between the US and the EU. President Donald Trump expressed his ire at American companies facing such heavy fines, and The Wall Street Journal alleged that the EU was using these fines in part as a bargaining chip in trade negotiations.

This article originally appeared on Engadget at https://www.engadget.com/big-tech/eu-tells-apple-it-has-no-intention-of-repealing-the-digital-markets-act-175950691.html?src=rss

Amazon is paying $2.5 billion to settle FTC claims it duped customers into signing up for Prime

Amazon will pay a record civil penalty to settle a case with the Federal Trade Commission. The agency accused Amazon of tricking consumers into signing up for a Prime membership without their consent and making it hard for customers to cancel in a lawsuit filed in 2023.

To settle the charges, Amazon has agreed to pay a $1 billion civil penalty and $1.5 billion to refund customers. The company also agreed to "ease unlawful enrollment and cancellation practices for Prime," per the FTC.

The agency says the civil penalty is the largest ever for a case involving a breach of its rules — it had accused Amazon of violating the FTC Act and the Restore Online Shoppers’ Confidence Act. The $1.5 billion in consumer redress will provide "full relief for the estimated 35 million consumers impacted by unwanted Prime enrollment or deferred cancellation," the FTC said. It added that this is the second-highest restitution award it has ever obtained.

Moreover, Amazon will be prohibited from using some of the dark patterns (i.e. deceptive design practices) it has employed to dissuade customers from canceling Prime memberships. For instance, the settlement precludes it from displaying a button that reads, "No, I don’t want Free Shipping" during the cancellation flow. Instead, it will have to show a "a clear and conspicuous button for customers to decline Prime," per the terms of the settlement, which does not include an admission of guilt on Amazon's part.

The company will also have to provide clearer information about a Prime subscription to consumers during the sign-up process. This will include details about the price, whether the subscription auto-renews and how to cancel.

"Amazon and our executives have always followed the law and this settlement allows us to move forward and focus on innovating for customers," Amazon spokesperson Mark Blafkin said in a statement provided to Engadget. "We work incredibly hard to make it clear and simple for customers to both sign up or cancel their Prime membership, and to offer substantial value for our many millions of loyal Prime members around the world. We will continue to do so, and look forward to what we’ll deliver for Prime members in the coming years."

Update September 25, 1:47PM ET: Added Amazon's statement and clarified that the settlement doesn't include an admission of guilt.

This article originally appeared on Engadget at https://www.engadget.com/big-tech/amazon-is-paying-25-billion-to-settle-ftc-claims-it-duped-customers-into-signing-up-for-prime-160641347.html?src=rss

Prime Day deals include this Shark AI Ultra robot vacuum for 58 percent off

Ahead of the Amazon Big Deal Days event (aka Prime Day in October), a tasty deal on a Shark robot vacuum has popped up. You'll need to be a Prime member to take advantage of the offer on the Shark AV2501S AI Ultra robot vacuum, but if you are, you can get the device for over half off. The discount drops the price from $550 to $230.

That means you can snap up the robot vacuum for $320 below list price. The discount marks a record low for this model.

Shark offers several variations of its AI Ultra robot vacuums. There are small variations between them, and a different model is our pick for the best robot vacuum for most people. In general, you can expect solid cleaning performance from these devices, along with accurate home mapping and an easy-to-use app.

The model that's on sale here is said to run for up to 120 minutes on a single charge, which should be enough to clean an entire floor in a typical home. The self-emptying, bagless vacuum can store up to 30 days worth of dirt and debris in its base. Shark says it can capture 99.97 percent of dust and allergens with the help of HEPA filtration.

If you'd rather plump for a model that's able to mop your floors too, you're in luck: a Shark Matrix Plus 2-in-1 vacuum is on sale as well. At $300 for Prime members, this vacuum is available for $400 (or 57 percent) off the list price. Its mopping function can scrub hard floors 100 times per minute. You can also trigger the Matrix Mop function in the app for a deeper clean. This delivers 50 percent better stain cleaning in targeted zones, according to Shark.

This article originally appeared on Engadget at https://www.engadget.com/deals/prime-day-deals-include-this-shark-ai-ultra-robot-vacuum-for-58-percent-off-171836638.html?src=rss

Microsoft blocks Israel's use of its data centers for mass surveillance of Palestinians

Microsoft has ended access to its data centers for a unit of the Israeli military that helped power a massive surveillance operation against Palestinian civilians, according to a report by The Guardian. The company says that the country's spy agency has violated its terms of service.

This surveillance system collected millions of phone calls made by Palestinian civilians every day in Gaza and the West Bank. The massive trove of data has been stored via Microsoft's Azure cloud platform, but the company just informed Israel's spy agency that this practice will no longer be acceptable. 

Microsoft’s vice-chair and president, Brad Smith, alerted staff of the move in an email, writing that the company had “ceased and disabled a set of services to a unit within the Israel ministry of defense." He went to suggest that this included cutting off access to cloud storage and some AI services.

“We do not provide technology to facilitate mass surveillance of civilians," he continued. "We have applied this principle in every country around the world, and we have insisted on it repeatedly for more than two decades.”

Microsoft came to this decision after conducting an external inquiry to review the spy agency's use of its Azure cloud platform. It also comes amid pressure from both employees and investors for the company to examine its relationship with Israel as it relates to the military offensive in Gaza.

This reportedly started back in 2021, when Microsoft CEO Satya Nadella allegedly okayed the storage effort personally after meeting with a commander from Israel's elite military surveillance corps, Unit 8200. Nadella reportedly gave the country a customized and segregated area within the Azure platform to store these phone calls, all without knowledge or consent from Palestinians. 

While conflict has existed between Israel and Palestinian groups for decades, these platforms were built out a full two years before the the most recent escalation in violence, beginning October 7, 2023. The mantra when building out the project was to record "a million calls an hour."

Leaked Microsoft files suggested that the lion's share of this data was being stored in Azure facilities in the Netherlands, but Israel allegedly moved it after Microsoft started its initial investigation. The Guardian has reported that Unit 8200 planned on transferring the data to the Amazon Web Services cloud platform. We have contacted Amazon to ask if it has accepted this gigantic trove of personal data. 

This article originally appeared on Engadget at https://www.engadget.com/big-tech/microsoft-blocks-israels-use-of-its-data-centers-for-mass-surveillance-of-palestinians-170107061.html?src=rss

Nintendo is opening a pop-up store in London this fall

Nintendo will open an official pop-up store in London this fall, its first-ever in the UK. The store will be open from October 22 to November 16 on the ground floor of the Westfield London shopping mall in Shepherd's Bush.

The Japanese gaming giant operates official stores in Tokyo and Kyoto, as well as permanent locations in New York and (20 years later) San Francisco. The pop-up store in London will feature an extensive collection of Nintendo merchandise across the brand's most iconic titles including Super Mario, Donkey Kong, The Legend of Zelda, Animal Crossing and more. Fans can pick up all sorts of apparel and collectible accessories like keyrings, pins and bags.

Entry to the pop-up during its first week will be by reservation only, with tickets going live on October 7. Guests will be able to reserve a date and time slot to enter between October 22 and 26. From October 27, store entry will operate via a first-come, first-served system, though don't be surprised to see a line out the door.

Microsoft pressured to extend free Windows 10 security updates in most of Europe

Score another win for EU consumers. On Wednesday, Microsoft backtracked on the strings it had attached to Windows 10's Extended Security Updates (ESU) program. However, it only did so in the European Economic Area (EEA) following complaints from an advocacy group. The elephant in the room? The EU's Digital Markets Act (DMA) safeguards against unlawful bundling and restricting consumer choice. Had the company not relented, those regulations could have spawned investigations and hefty fines.

"In the European Economic Area, we’re making updates to the enrollment process to ensure it meets local expectations and delivers a secure, streamlined experience," a Microsoft spokesperson told Windows Central. "Our goal is to support customers and provide them with options as they transition to Windows 11, with uninterrupted access to critical security updates."

The move stems from Microsoft's decisions regarding the future of Windows 10. In early 2023, the company said it would only support crucial security updates for the operating system until October 2025.

Microsoft

Since then, the company launched the ESU program. It allows consumers to pay $30 to add a one-year extension on security updates. The Windows maker also added a couple of "free" options. The first one is to enable Windows cloud backup using a Microsoft account. However, that uses OneDrive. And once you surpass 5GB of free storage, you'll need to pay for more. (If you have a lot of big files to back up, that would likely apply.) The other option is to spend 1,000 Microsoft Rewards points, which not everyone has lying around.

The advocacy org Euroconsumers called out Microsoft in June. "Limiting security means limiting product life," the nonprofit wrote. It described Microsoft's Windows 10 requirements as "planned obsolescence." That's the practice of deliberately designing products with limited lifespans. Conveniently for Big Tech, that tends to boost sales. Inconveniently for consumers, it's an added cost.

Planned obsolescence is also a term that could have drawn the ire of EU regulators. Tech titans like Microsoft found to violate the DMA can be forced to pay up to 10 percent of their global annual turnover.

Euroconsumers said Microsoft was steering consumers toward two options. "Buy a new device before you're ready or stick with your current one and face increased security risks," it wrote. The group noted that 22 percent of EU PC owners still run a Windows device from 2017 or earlier. 2018 is often considered a rough baseline for Windows 11 eligibility, given the software's technical requirements.

While PC users in the EEA now have an extra year, the rest of the world isn't so lucky. If you're still running Windows 10, you only have until October 14 to do one of three things: upgrade to Windows 11 (Microsoft's preferred choice), pay $30, turn on Windows backup, or fork over reward points. Choose wisely.

This article originally appeared on Engadget at https://www.engadget.com/computing/microsoft-pressured-to-extend-free-windows-10-security-updates-in-most-of-europe-164533056.html?src=rss

Elon Musk’s Grok is cleared for federal government use

Despite Elon Musk and Donald Trump’s very public spats and seemingly still ongoing feud, the White House has remained committed to supporting the former’s AI ambitions. And today, the General Services Administration (GSA) announced that it has reached an agreement with xAI that will allow it to buy Musk’s Grok AI models for $0.42 per organization.

As part of the Trump administration’s OneGov procurement initiative, the deal with xAI will allow federal agencies access to the Grok 4 and Grok 4 Fast advanced reasoning models. The 18-month contract is the longest OneGov AI procurement agreement to date. xAI announced its Grok for Government strategy earlier in the summer, which signalled its intention to provide the government with a suite of AI products, including custom models for national security, science and healthcare purposes.

As well as opening its models for government use, xAI is also providing dedicated engineers to speed up the implementation of its AI tools for participating agencies, and will offer an "upgrade path" for expanded features and higher rate limits. Such access is a crucial part of Trump’s AI Action Plan, designed to position the US as the global leader in AI. And his administration doesn’t appear to have been put off by Grok’s bizarre behavior in recent months, such as its preoccupation with far-right conspiracy theories regarding "white genocide" in South Africa, or its brief but enthusiastic turn towards antisemitism.

xAI is the latest in a line of AI companies to strike deals with the GSA. Back in August, Anthropic began offering its Claude AI model to three branches of the US government for $1, following Gemini and xAI’s arch rival OpenAI joining a list of approved vendors.

This article originally appeared on Engadget at https://www.engadget.com/ai/elon-musks-grok-is-cleared-for-federal-government-use-162407911.html?src=rss

Nothing is spinning off its budget CMF brand

Nothing is spinning off its budget brand CMF into an independent subsidiary, according to a report by TechCrunch. India will serve as the newly-formed company's headquarters for manufacturing and R&D, thanks to a partnership with one of the country's telecom operators. Nothing says this venture will create over 1,800 jobs in the country.

India is actually a good choice for the HQ. The country is Nothing's strongest market overall, with over a 2 percent market share in smartphones. It's the fastest-growing brand in the region, with an 85 percent growth in shipments year over year. CMF also specializes in budget-friendly handsets under $200 and this is the dominant category in India, according to the IDC.

“India will play a key role in shaping the future of the global smartphone industry. CMF has been well-received by the market since we launched it two years ago," Nothing CEO Carl Pei said in a statement. "With our end-to-end capabilities, we are uniquely positioned to now build it into India’s first truly global smartphone brand."

Nothing hasn't said how much money it would take to set up this new subsidiary, but the company did just raise $200 million in a funding round. Maybe some of that cash will be funneled to this project.

The company first launched CMF back in 2023. It started with the release of a pair of earbuds and a smartwatch but has since moved on to smartphones.

This article originally appeared on Engadget at https://www.engadget.com/mobile/smartphones/nothing-is-spinning-off-its-budget-cmf-brand-152505010.html?src=rss

Spotify is doing more to address AI 'slop' on its platform

Spotify has announced a set of policy changes surrounding AI-generated music and spam on its streaming platform. The company is helping to develop an industry standard for AI disclosure in music credits, alongside DDEX. It will be strengthening its approach to AI-assisted spam, such as unauthorized vocal clones, as well as uploaded music that fraudulently delivers music to another artist's profile.

The new disclosures will encourage artists to share what aspect, if any, of their production was created with the assistance of AI. Instead of a song simply being marked as "is AI" or "no AI," artists can specify whether they used AI-generated vocals, instrumentation or post-production.

The streamer will also debut a new impersonation policy, making it clearer how the platform deals with AI voice clones. The policy promises to give artists stronger protections against this sort of spam, and clearer recourse should any appear.

"…the pace of recent advances in generative AI technology has felt quick and at times unsettling, especially for creatives. At its best, AI is unlocking incredible new ways for artists to create music and for listeners to discover it. At its worst, AI can be used by bad actors and content farms to confuse or deceive listeners, push 'slop' into the ecosystem, and interfere with authentic artists working to build their careers," Spotify said in its announcement.

These aren't the only tactics that bad actors use to divert royalties and deceive listeners. Spotify shared that other types of spam "such as mass uploads, duplicates, SEO hacks, artificially short track abuse, and other forms of slop" have become easier to create and deploy as AI tools substantially lower the barrier of entry to creating this type of content.

To address these, the streamer is launching a new spam filter this fall that will identify uploads and tracks that engage in these types of spam, tag them on the platform, and stop recommending them to users. Spotify said that over the past 12 months it has already removed more than 75 million "spammy" tracks.

Spotify says that this sort of spam can dilute the royalty pool and take attention away from real artists trying to earn a living, even in part on the platform. The company says its goal is to achieve more transparency for listeners and protect artist identity through these new policies.

These new policies don't address AI-generated projects like The Velvet Sundown, which remains on the platform despite all its lyrics, vocals, and imagery being entirely AI-generated. Spotify doesn't directly acknowledge the AI band, but says "we support artists’ freedom to use AI creatively while actively combating its misuse by content farms and bad actors."

This article originally appeared on Engadget at https://www.engadget.com/entertainment/streaming/spotify-is-doing-more-to-address-ai-slop-on-its-platform-151102801.html?src=rss

The best October Prime Day deals to get today: Early sales on tech from Apple, Roku, Shark, Anker and more

Now that we know October Prime Day is on the horizon, it’s time to start thinking about what you may want to snag at a discount during the sale. If you pay the $139 annual fee for Prime, sale events like these are a great time to stock up on essentials and cross things off your wishlist while you can save some money.

Most discounts will be exclusively available to Prime subscribers, but there are always a few that anyone shopping on Amazon can grab. Similarly, there are always early deals in the days and weeks leading up to Prime Day, and this year is no different. Here, we’ve collected the best October Prime Day deals you can shop for right now and we’ll keep updating this post as we get close to Prime Day proper.

Apple iPad (A16) for $299 ($50 off): The new base-model iPad now comes with twice the storage of the previous model and the A16 chip. That makes the most affordable iPad faster and more capable, but still isn't enough to support Apple Intelligence.

Apple Mac mini (M4) for $499 $100 off): If you prefer desktops, the upgraded M4 Mac mini is one that won’t take up too much space, but will provide a ton of power at the same time. Not only does it come with an M4 chipset, but it also includes 16GB of RAM in the base model, plus front-facing USB-C and headphone ports for easier access.

Apple iPad Air (11-inch, M3) for $449 ($150 off): The only major difference between the latest iPad Air and the previous generation is the addition of the faster M3 chip. We awarded the new slab an 89 in our review, appreciating the fact that the M3 chip was about 16 percent faster in benchmark tests than the M2. This is the iPad to get if you want a reasonable amount of productivity out of an iPad that's more affordable than the Pro models.

Jisulife Life7 handheld fan for $25 (14 percent off, Prime exclusive): This handy little fan is a must-have if you life in a warm climate or have a tropical vacation planned anytime soon. It can be used as a table or handheld fan and even be worn around the neck so you don't have to hold it at all. Its 5,000 mAh battery allows it to last hours on a single charge, and the small display in the middle of the fan's blades show its remaining battery level.

Roku Streaming Stick Plus 2025 for $29 (27 percent off): Roku makes some of the best streaming devices available, and this small dongle gives you access to a ton of free content plus all the other streaming services you could ask for: Netflix, Prime Video, Disney+, HBO Max and many more.

Anker 622 5K magnetic power bank with stand for $34 (29 percent off, Prime exclusive): This 0.5-inch thick power bank attaches magnetically to iPhones and won't get in your way when you're using your phone. It also has a built-in stand so you can watch videos, make FaceTime calls and more hands-free while your phone is powering up.

Leebein 2025 electric spin scrubber for $40 (43 percent off, Prime exclusive): This is an updated version of my beloved Leebein electric scrubber, which has made cleaning my shower easier than ever before. It comes with seven brush heads so you can use it to clean all kinds of surfaces, and its adjustable arm length makes it easier to clean hard-to-reach spots. It's IPX7 waterproof and recharges via USB-C.

Anker Nano 5K ultra-slim power bank (Qi2, 15W) for $46 (16 percent off): A top pick in our guide to the best MagSafe power banks, this super-slim battery is great for anyone who wants the convenient of extra power without the bulk. We found its proportions work very well with iPhones, and its smooth, matte texture and solid build quality make it feel premium.

Shark AI robot vacuum with self-empty base for $230 (58 percent off, Prime exclusive): A version of one of our favorite robot vacuums, this Shark machine has strong suction power and supports home mapping. The Shark mobile app lets you set cleaning schedules, and the self-empty base that it comes with will hold 30 days worth of dust and debris.

Levoit LVAC-300 cordless vacuum for $250 ($100 off, Prime exclusive): One of our favorite cordless vacuums, this Levoit machine has great handling, strong suction power for its price and a premium-feeling design. Its bin isn't too small, it has HEPA filtration and its battery life should be more than enough for you to clean your whole home many times over before it needs a recharge.

Samsung EVO Select microSD card (256GB) for $23 (15 percent off): This Samsung card has been one of our recommended models for a long time. It's a no-frills microSD card that, while not the fastest, will be perfectly capable in most devices where you're just looking for simple, expanded storage.

JBL Go 4 portable speaker for $40 (20 percent off): The Go 4 is a handy little Bluetooth speaker that you can take anywhere you go thanks to its small, IP67-rated design and built-in carrying loop. It'll get seven hours of playtime on a single charge, and you can pair two together for stereo sound.

Anker MagGo 10K power bank (Qi2, 15W) for $63 (22 percent off, Prime exclusive): A 10K power bank like this is ideal if you want to be able to recharge your phone at least once fully and have extra power to spare. This one is also Qi2 compatible, providing up to 15W of power to supported phones.

Rode Wireless Go III for $199 (30 percent off): A top pick in our guide to the best wireless microphones, the Wireless Go III records pro-grade sound and has handy extras like onboard storage, 32-bit float and universal compatibility with iPhones, Android, cameras and PCs.

Shark Robot Vacuum and Mop Combo for $300 (57 percent off, Prime exclusive): If you're looking for an autonomous dirt-sucker that can also mop, this is a good option. It has a mopping pad and water reservoir built in, and it supports home mapping as well. Its self-emptying base can hold up to 60 days worth of debris, too.

Nintendo Switch 2 for $449: While not technically a discount, it's worth mentioning that the Switch 2 and the Mario Kart Switch 2 bundle are both available at Amazon now, no invitation required. Amazon only listed the new console for the first time in July after being left out of the initial pre-order/availability window in April. Once it became available, Amazon customers looking to buy the Switch 2 had to sign up to receive an invitation to do so. Now, that extra step has been removed and anyone can purchase the Switch 2 on Amazon.

This article originally appeared on Engadget at https://www.engadget.com/deals/the-best-october-prime-day-deals-to-get-today-early-sales-on-tech-from-apple-roku-shark-anker-and-more-050801467.html?src=rss

Forza Horizon 6 takes the arcade racing series to Japan in 2026

Microsoft has officially unveiled the next Forza Horizon game, confirming months of rumors that the latest entry in its consistently excellent open-world racing series will be set in Japan. Announced, rather fittingly, at Tokyo Game Show in an Xbox briefing, Forza Horizon 6 has a 2026 release date and will launch first on Xbox and PC. A few years ago it would have stayed on those platforms, but after Forza Horizon 5 made its way to PS5 earlier this year, all bets are off on Microsoft exclusivity in the future.

As one of Xbox’s most important series, it’s a bit disappointing that we didn’t get any gameplay in the TGS teaser trailer. Instead, the camera pans across what looks like a workbench, with license plates and trinkets from countries featured in previous entries hung on the wall behind it. Eventually we get to Japan, and the camera moves up to reveal Mount Fuji in the distance, with the obligatory cherry blossom in the corners of the frame.

What the announcement lacked in gameplay, though, it somewhat made up for with the accompanying Xbox Wire post, which sheds some light on Playground Games’ decision to choose the Land of the Rising Sun as the location for Forza Horizon 6. According to the game’s Art Director, Don Arceta, the "unique culture" of Japan — from its cars, to its music and fashion — was a driving factor, and added that with five Horizon games already under its belt, the studio feels it can build an "authentic representation" of the country that does it justice.

Playground Games is keeping its cards close to its chest where real-life locations featured in the game are concerned, but unsurprisingly confirmed that Tokyo is one of them, as well as some of the country’s rural and mountain areas. The studio isn’t saying much about cars yet, either, but promised that Japanese car culture will be reflected. It also confirmed that seasons will once again be a big part of the experience.

Forza Horizon 5 came out in 2021 and turned Engadget’s "not much of a car guy" Nathan Ingraham into a racing game fan with its incredibly detailed Mexico setting, approachable driving and accessible open-world design.

This article originally appeared on Engadget at https://www.engadget.com/gaming/forza-horizon-6-takes-the-arcade-racing-series-to-japan-in-2026-142253601.html?src=rss

Surfshark VPN review: A fast VPN for casual users

Surfshark is one of the youngest major VPNs, but it's grown rapidly over the last seven years. Since 2018, it's expanded its network to 100 countries, added a suite of apps to its Surfshark One package and experimented with advanced touches like servers that constantly rotate your IP address.

Formerly a budget VPN, Surfshark is still pretty affordable but the cost of a one-month subscription is way too high. These days, it's more interested in innovation than affordability. Its features may not always be cutting-edge — Nexus is basically Tor and Everlink is a slightly better kill switch — but it's rare to find either implemented so well in a commercial VPN. Plus, download and upload speeds are fantastic.

It all adds up to one of the best VPNs for casual users, even when compared to others in its weight class. But those already deeply familiar with VPNs might trip over some aspects. For this review, I studied 11 different angles on Surfshark's service, some through hands-on tests and others with old-fashioned journalism. Check out the sections most important to you, then head to the end for my final verdict.

Editor's note (9/25/25): We've overhauled our VPN coverage to provide more detailed, actionable buying advice. Going forward, we'll continue to update both our best VPN list and individual reviews (like this one) as circumstances change. Most recently, we added official scores to all of our VPN reviews. Check out how we test VPNs to learn more about the new standards we're using.

• Findings at a glance

• Installing, configuring and using Surfshark

• Surfshark speed test: Super fast with some latency

• Surfshark security test: Simple but effective protocols

• How much does Surfshark cost?

• Surfshark side apps and bundles

• Close-reading Surfshark's privacy policy

• Can Surfshark change your virtual location?

• Investigating Surfshark's server network

• Extra features of Surfshark

• Surfshark customer support options

• Surfshark background check: No real blemishes

• Final verdict

This table summarizes everything I discovered about Surfshark. For details on any bullet point, see its section in the full review.

For this section, I'll be going over how it feels to use Surfshark. I'll look at how easy or difficult it is to install, what you have to do to set it up and whether the user interface (UI) helps out or gets in the way on each platform.

Windows

Surfshark is easy to download and install on Windows. You can get it through the Microsoft store, but going directly through the Surfshark website makes sure you have the latest and most complete version. The downloader asks you to click "Yes," but mostly handles itself, and finishes up in a short time.

Sam Chapman for Engadget

The first point that might divide casual users and VPN veterans comes once you open the app for the first time. Surfshark insists on taking you through a guided setup process for all its features. This is extremely helpful if you've never used a VPN before. If you have, its repeated insistence on sending you through the tour can get aggravating.

This pattern repeated itself as I went through the app. The front page is sensibly organized and arranged so I could see all the facts and settings — but it remains way too reliant on unskippable pop-ups to provide information. Messages appeared when I tried to cancel a stuck connection, when I turned settings on and off and even when I tried to close the app. It's usually important information, but displaying a message that freezes every other button is a bad move.

Sam Chapman for Engadget

It's a shame, because everything else works fine. The settings tabs do an admirable job of packing multiple, semi-related apps into a single UI. Just take note that a lot of stuff is enabled by default, including auto-connect and NoBorders, which may make the VPN behave in ways you don't need.

Mac

Surfshark's app for macOS is mostly the same as its Windows app, for better or worse. It's got the same split-second download and installation and the same clear organization of tabs with well-explained features. It's also got the same relentless popups, settings on by default and occasionally sticky connections. If it gets hung up at 95 percent, just wait — it's still working.

Sam Chapman for Engadget

Android

The Surfshark Android app, available through the Google Play Store, is ready after you simply download and login. It doesn't reinvent the wheel, but Surfshark's design language turns out to be very easy to compress into mobile — except that it's dark-mode by default, while all the other three are light. I found everything I needed within moments. Android even removes almost all the full-screen stop-everything notifications.

Sam Chapman for Engadget

One thing to be aware of: when you start up the Android app, the VPN is visible, but the landing page includes all the features of Surfshark One. To get the VPN-only front page, click the arrow across from "VPN" at the top of the screen. That said, you can do almost everything you need to do no matter which page you're on.

iOS

It's unusual to see a big difference between a VPN's apps on the two mobile platforms, but Surfshark switches up its design from Android to iOS. Everything now lives in four tabs on the bottom of the window. The settings tab is great, largely because it looks like how preferences do on Mac and Windows.

Sam Chapman for Engadget

The main VPN window is a bit jankier. It works fine, but in order to make the server list constantly visible (unnecessary IMO), it relegates the actual VPN connection to a floating box that looks like an afterthought. It even includes a button to open the server list that's already open. I freely admit this is just a personal complaint, but it still baffles me.

Browser extensions

Surfshark's browser extensions for Firefox and Chromium are almost clones of its desktop apps, with a few differences. The Alert tab tells you if the site you're on has suffered a security breach recently. CleanWeb 2.0 is the main reason to add Surfshark to your browser; it's a much stronger blocker that even works on some video insert ads.

Sam Chapman for Engadget

All VPNs add to your web browser's workload, so it's normal to see speeds drop. The best VPNs manage to keep the drops as small as possible. I used speedtest.net to see how quickly my download speeds declined at six of Surfshark's server locations — that's a measure of how fast a browser can load web pages, buffer videos and download large files.

I also looked for latency, which affects real-time communication with another server (important for gamers) and upload speed, which affects how quickly your device can send data (important for torrenting). My results are in the table.

In terms of download and upload speeds, Surfshark is the fastest VPN I've tested so far, beating out ExpressVPN, Proton VPN and NordVPN — and none of those are exactly sluggish. If you look at the Indonesia row, you'll see I actually got a higher upload speed than I did without the VPN active. That probably has more to do with natural fluctuations from my ISP than anything Surfshark did, but it still indicates how little this VPN weighs down your browsing.

Sam Chapman for Engadget

Latency is more of a mixed bag. Jumps in ping time are expected across long distances, but I also saw a fairly large jump on a server in the same state as me. Surfshark may not be ideal for fast-paced online games, but its speeds are just about perfect for any other task.

To determine whether Surfshark can keep you safe, I'll be looking at three factors: whether it uses secure and up-to-date VPN protocols, whether any of those protocols leak your IP address and whether encryption is being applied to all data packets. The first is a matter of researching the technology, but I've run practical tests for the other two.

VPN protocols

Surfshark uses the VPN protocols WireGuard, OpenVPN (over TCP or UDP) and IKEv2. All of these are field-tested protocols that rely on uncracked encryption and proven authentication methods. There's no proprietary protocol with shady source code, nor any outdated legacy options that might pose a risk if you pick them. My only complaint is that OpenVPN isn't available on Mac.

Sam Chapman for Engadget

Each of these protocols does the same thing: establish communication between your device, VPN servers and the internet. They handle the steps differently, but the average user is only likely to see small differences. If one protocol isn't working, you can switch to another — or just pick the automatic protocol selection option, which worked fine during my tests. If you want to pick for yourself, use the table below for reference.

Leak test

Next, I picked five of the servers I didn't use for the speed test and checked all of them for three different kinds of leaks using ipleak.net. Surfshark has its own DNS servers, so DNS leaks are unlikely — those usually come from a VPN using public servers to resolve DNS requests. As expected, DNS requests didn't reveal my IP address in any of the tests.

Sam Chapman for Engadget

There's more potential for IPv6 leaks, as Surfshark doesn't support IPv6. It recommends turning it off instead, and has guides on doing that. To fully put Surfshark through its paces, though, I ran my tests with IPv6 still active on my computer. On all five servers, the IPv6 test could not run. This indicates that Surfshark actively blocks IPv6, which (until more of the internet starts running on the new protocol) is a fine way to prevent IPv6 leaks.

The leak testing tool also checks for WebRTC leaks, so I ran each test with a Google Meet call active, but that didn't leak either. In short, Surfshark didn't spring a leak no matter how many times I poked it. I'm prepared to call it watertight.

Encryption test

To be absolutely sure I was recommending a safe VPN, I used WireShark (no relation) and HTTP Forever to see whether its encryption works. From examining the data stream before and after connecting to a Surfshark server, it's clear that encryption is truly being applied. Take a look for yourself — the fact that you can't get any information from the screenshot means the VPN protocol is working.

Sam Chapman for Engadget

Surfshark is part of a suite of security products — a lot like its sister brand NordVPN, but with the offerings per tier a little more coherent. It comes in three subscription types, each of which is available at three durations. The table below shows the whole range of costs.

Note that the 24-month plans are introductory only. You can only renew for one month or 12 months. The prices below also factor in extra months you'll get when you sign up for the first time, which aren't available in renewals.

Surfshark's price drops steeply from its shortest to its longest subscriptions. $15.45 is way overpriced (even for a service we quite like), but $1.99 per month is about the least you'll pay for any VPN worth recommending. The jump from Starter (which only includes the VPN service) to One (which includes everything else except Incogni) is also tiny on the 12-month plan — only an extra $0.20.

Surfshark is almost certainly doing this to juice active user numbers — but that doesn't mean you can't take advantage of it. To our mind, the 12-month subscription to Surfshark One is extremely good value, provided you already know you want Surfshark VPN.

Every plan comes with a 30-day money-back guarantee. At any point before 30 days have elapsed, you can start a live chat conversation to get your money back. This is only allowed once every six months.

You can also get a free trial for seven days by installing the Surfshark app from the Apple or Android app store, creating an account, then choosing the plan that includes the 7-day free trial. However, since you have to enter a payment method for this — and it's not accessible on Windows — the 30-day period is generally more convenient.

Here's what you can get from Surfshark aside from its VPN service. In order to keep our focus on the VPN, we didn't have time to rigorously test all these, but we've included the basic information to help you make your decision.

Alternative ID

Alternative ID is included along with VPN service in a Surfshark Starter plan. It consists of two features: masked email generator and personal detail generator.

Masked email generator gives you a random email address that sends all messages along to your real inbox. Since it auto-forwards everything, an email mask isn't for reducing spam. Instead, it lets you keep your real email address secret when signing up for online services — since providing your actual address runs the risk of leaking it. For an extra $2.89 per month, you can also get an alternative phone number that works similarly.

Sam Chapman for Engadget

Personal detail generator creates fake names, addresses and birthdays you can use to sign up for any websites you're uncertain about. Be sure to use it along with the VPN, or your IP address will reveal the actual details. It also might be helpful for randomly generating your next RPG character.

Surfshark Antivirus

Surfshark Antivirus is included in Surfshark One and One+ subscriptions. It can scan files as you download them and can also scan your whole system for malware, checking against a list of unwanted programs it claims to update every three hours.

AV-Test gave it a perfect score in protection. A top score (6 out of 6) means Surfshark Antivirus caught 100 percent of malware thrown at it, both known threats and zero-day attacks. It also got a perfect usability score for totally avoiding false positives, and a 5.5 out of 6 on performance — it had nearly no impact on downloading apps or copying files, but slowed down popular websites by around 30 percent on average.

Surfshark Alert

Surfshark Alert, which is also part of Surfshark One and One+, scans the information released in data breaches to see if any of your important data has been compromised. It can check for email addresses, usernames, passwords, credit card numbers and government IDs. It can also warn you when malware on your computer is capable of stealing your data.

Surfshark Search

You can get access to Surfshark Search, a search engine akin to Google, with a One or One+ subscription. It offers a number of benefits over Google, including being totally ad-free (a perk of being subsidized by subscription). You can also change the region from which you get results. Surfshark claims that all results are totally organic, with no trackers or learning from logged searches.

Incogni

Incogni, an automated data removal service, is the one feature restricted to Surfshark One+ plans. It searches data brokerages to see if they have any of your information, then contacts them for you to request they purge the data. Brokers don't have to honor all deletion requests, but it never hurts to try. Incogni keeps searching as long as your subscription is active, so it can automatically send multiple requests to the same broker if necessary.

Saily eSIM

Several VPNs now provide their users with Saily eSIM discounts, which can be used to get regular phone service in foreign countries. Surfshark grants various discounts through the coupon code surfshark5, which appears to be usable even without a Surfshark account.

Surfshark's privacy policy shows us its outward attitude toward user privacy. While it's possible for a VPN provider to lie outright in its posted policy, they prefer not to for liability reasons. An unreliable service will instead use loopholes, vagueries and incomplete statements to make privacy promises it doesn't plan on keeping.

By contrast, a VPN service worth your money will be as specific as possible about any exceptions to its no-logs policy. I read Surfshark's entire privacy policy to see which type of VPN it is — no small feat, as it's well over 5,000 words long. The page includes summaries, but they're pared down a little too far to be useful, so it's into the full text I go.

Sam Chapman for Engadget

Surfshark is based in the Netherlands, which makes it subject to the European Union's General Data Protection Regulation (GDPR). That's good news, since GDPR does not permit any company to retain data without justifying it. The Netherlands is likely part of the Nine Eyes intelligence-sharing agreement, but agencies can't spy on a VPN's users unless that VPN is already saving activity logs — and if it is, it's got much bigger problems than where it's based.

Surfshark's entire server network runs on RAM-only infrastructure, so any saved activity is erased at regular intervals. Even better, the policy is clear that your activity on any server is deleted within 15 minutes. If we take Surfshark at its word, there's very little reason to worry about servers being seized or searched.

That means the only risk is that Surfshark itself might take advantage of its access to your information. It does, unfortunately, claim the right to view information gathered by third-party trackers, such as the cookies used by social media sites to track ad performance. But none of those trackers are watching what you do when you use the VPN, nor digging into your actual browsing activity — which, as we've seen, doesn't exist after 15 minutes.

Independent privacy audits

Although we can't conclusively verify any VPN's privacy policy, since that involves the logically impossible feat of proving a negative, we can get very close by checking independent audits. Third-party auditors can verify a VPN's public statements through a combination of interviews, document review and directly tinkering with the infrastructure.

Surfshark passed its most recent audit in June 2025, which — at time of writing — leaves very little room for interference. Deloitte, the audit firm, concluded that "the configuration of IT systems and management of the supporting IT operations is properly prepared, in all material respects in accordance with the Surfshark’s description set out in the Appendix I."

I checked that description and found that it matches up to Surfshark's public privacy policy. So, as much as we can say that any VPN is private, Surfshark is private. It may not stop all personalized ads, but it will at least keep your personal information secret.

If you mainly need a VPN for streaming, this is the section for you. Here, I test whether Surfshark can change your location convincingly enough for you to see content from other countries, without getting caught and blocked by streaming platforms.

Sam Chapman for Engadget

The procedure is simple: I pick five test locations, connect to three different servers in each and load Netflix. If I get in cleanly without error messages, and actually see different shows than I would without a VPN, I'll know Surfshark convincingly changed my location. (As a reminder, when unblocking with any VPN, always make sure you're logged into your streaming account first.)

Surfshark performed extremely well in this test, not only unblocking every location at least twice but also maintaining fast enough speeds to keep the videos from stuttering. The only failure was that one server in Japan got spied out as a VPN and blocked. However, simply disconnecting and reconnecting was enough to get me over the hurdle.

Sam Chapman for Engadget

Any failure to change virtual location counts against a VPN, but even the best services can't keep their servers off 100 percent of blocklists. NordVPN, another top-five VPN, scored the same 14/15. As far as I'm concerned, that's a passing grade for streaming — but if it's a dealbreaker for you, remember ExpressVPN and Proton VPN both have perfect scores.

Surfshark has a total of 100 server locations in 141 countries. Of those locations, 47 are virtual and 92 are physical. A virtual VPN server gives you an apparent digital location different from where it's really located. There's little security tradeoff, but speeds may vary depending on how far you are from the server's actual position. The one count against Surfshark here is that it doesn't clarify where its virtual servers are really located, so using them is a speed crapshoot.

The table below shows the proportions of server locations in each region, so you can decide whether Surfshark is likely to be useful for where you are. You can find a full list on Surfshark's website, though only the apps show which servers are actually virtual.

These numbers are interesting in a few ways. First, Surfshark's server network is only about one-third virtual, the lowest proportion of any major VPN I've reviewed so far. There's also no continent without a single physical server. This makes Surfshark a very good choice to guarantee usable internet speeds while globetrotting.

As usual, servers are mostly in Europe and North America, although there's a fair amount of Asia represented by the virtual locations. Surfshark is newer than most of its competitors, so it's impressive that it's built such an extensive network in half the time. The table reveals a VPN service that's interested in expanding, but won't do so at the expense of establishing quality locations that actually work.

Surfshark sets itself apart from the VPN pack with several features, some visible to the user and some working behind the scenes. Mostly, its features are familiar concepts implemented well, rather than outright innovations like NordVPN's Meshnet — but there's a lot of possibility for exciting new directions.

Underlying technology: Surfshark Nexus and Everlink

Surfshark has two technological innovations under the hood that make its VPN relatively unique. The first is Nexus, which has been active since 2022. Nexus connects all Surfshark's servers together into a tighter network than most VPNs use. All its servers are nodes in a web of interconnected paths, a lot like the Tor network, except proprietary to one company.

When you choose a server location, you're only selecting the last node in a sequence. Surfshark software selects the entry point and the path to your chosen exit. This gives every user more options for a working connection path, and allows Surfshark to regularly rotate your apparent IP address so activity on one masked identity can't trace back to you. It's also what makes it possible to pick the two steps in a double-hop connection (see "Dynamic MultiHop" below).

Sam Chapman for Engadget

Everlink is a newer technology that Surfshark calls a "self-healing VPN infrastructure." Any time you're connected to Surfshark through the WireGuard protocol, you're also connected to Everlink — a parallel infrastructure that watches your VPN connection. If it ever drops, Everlink instantly reconnects you to another server nearby, theoretically so fast you won't notice the outage.

This is basically one step up from a kill switch. Instead of cutting off your connection, Everlink switches it to the next server instead. This makes it difficult to test — the way to check a kill switch is to simulate a server drop by cutting off your own internet connection, which makes Everlink's function impossible. However, I can at least say I'm cautiously excited about Everlink's potential power to negate server drops.

Unlimited simultaneous connections

One thing you'll regularly hear about Surfshark is that it puts no limits on how many devices can connect through your account. In theory, one subscription is enough for you, your household and any friends that want to borrow it.

In practice, though, "unlimited" is a stretch. A support article acknowledges that some people may abuse the privilege by reselling access, or by masking illegal botnet activity behind hundreds of Surfshark connections.

The article doesn't explicitly say Surfshark will limit or ban abusers, but questions to the support team revealed that an account could be locked if an automated system detects fraud. If you use a high number of simultaneous connections, even legitimately, you could be flagged — though you may be able to unlock your account by telling customer service what happened.

Bypasser split tunneling

Bypasser lets you split your internet connection so certain apps or websites go through the VPN while others remain outside. This gets you better speeds on sites and apps that don't involve any sensitive information. Also, whenever a service requires your real IP address, you can keep using the VPN for everything else. Torrenting is a classic use case — you can keep BitTorrent protected in the background while you do something innocuous without VPN protection, improving speeds for both processes.

Sam Chapman for Engadget

Bypasser split tunneling is available on all of Surfshark's desktop and mobile clients, though iOS users can only split by website, not by app. Windows and Android users can also choose between two forms of split: "Bypass VPN," which keeps all listed sites and apps outside the VPN tunnel, and "Route via VPN," in which only the listed sites and apps stay within the VPN tunnel.

Dynamic MultiHop

Many VPNs have a double-hop option, which routes connections through two VPN servers to keep you protected in case one breaks. However, most of them restrict you to set paths of two servers. Surfshark is the only VPN that lets you choose both steps, thanks to Nexus pre-establishing paths between all the servers.

Sam Chapman for Engadget

For the best performance, I recommend picking an entry node near your real location, then choosing your exit node based on what you need to unblock. If you aren't unblocking anything in particular, your best bet is to simply choose two nodes in your own country.

Camouflage Mode obfuscation

VPN obfuscation refers to VPNs hiding the fact that they're VPNs at all — so not only is your IP address not your real one, but it doesn't appear to have been changed. Surfshark's obfuscation is called Camouflage Mode. As long as you're connected via OpenVPN, it's active automatically. So, if you suspect an outside firewall is preventing you from getting online while running Surfshark, switching to OpenVPN might solve the problem.

NoBorders mode

NoBorders is another option for getting online with Surfshark under external restrictions, especially those imposed by entire governments (like the Great Firewall of China). If Surfshark detects restrictions on internet usage, NoBorders activates and switches you to a server well-placed to get around those restrictions — for example, an IP address from the next country over.

Sam Chapman for Engadget

The problem with NoBorders is that it's active by default, which may restrict you to certain servers when you don't need the help. I recommend switching NoBorders off in the settings for better performance. It may turn on again if it detects possible restrictions, so check the setting again if you find your speeds slower than usual.

CleanWeb

CleanWeb, Surfshark's ad blocker, is available in two forms. The basic version of CleanWeb, which blocks banner ads and trackers, comes with all Surfshark apps.

Sam Chapman for Engadget

CleanWeb 2.0 comes as a browser extension, which means it only works in-browser. However, it's a much stronger blocker, capable of cancelling video ads on streaming and YouTube. It also automatically rejects non-vital cookies on every site so you don't have to click the button — which I appreciated, as my ADHD brain is easily distracted by pop-ups.

Static and dedicated IPs

Surfshark gives you a list of static IP addresses at no extra charge, another rarity among leading VPNs. A static IP means you'll have the same address every time you connect to the VPN, so you won't look suspicious for regularly opening the same service with a different IP. Without this, you'll find yourself hitting CAPTCHAs more often. You can also keep your home device connected to a static IP server and use that to access your home network remotely.

For an extra cost, Surfshark also offers dedicated IPs, which are the same as static IPs except you're the only person using them. Sometimes, a static IP can get blocked because of one bad user's behavior; a private dedicated IP removes that risk.

Other features

A few other features are worth noting. Surfshark has a kill switch that cuts off your internet connection if your VPN server ever drops — this is largely obsolete with Everlink, but remains in the background as an extra precaution. You can set the VPN to connect automatically except on trusted networks, and can choose the location to which it auto-connects.

Sam Chapman for Engadget

If you need to briefly work without the VPN, you can pause Surfshark for five minutes, 30 minutes or two hours. Finally, Android users can have their VPN server override their GPS location, moving it to wherever their IP address is currently. You obviously don't want to use this while your phone is navigating you somewhere, but otherwise it's useful for privacy.

Surfshark includes a few links to troubleshooting pages directly in the app, a tantalizingly unfinished feature. If you go to the settings tab and click Get Help and then Browse Guides, you'll see several links to FAQ articles about the version of the VPN client you're using. Clicking any link opens the page in your default browser. Jumping directly from the app to the relevant guide is really convenient, but if your problem isn't one of the five most common, you're out of luck — just click More Guides to go to the main support page.

Sam Chapman for Engadget

Once you're here, you'll find a helpful set of articles that make good design choices to put the average user at ease. "Getting started" includes setup guides for every Surfshark app, with abundant screenshots. Help articles are effectively written from the user's perspective. Instead of technical descriptions of problems, they start with topics like "I can't connect to Surfshark" or "I'm getting an error," then provide a sequence of DIY solutions.

The only category without a clear direction is "Surfschool," which gathers a disparate bunch of feature explanations, advanced tips and basic VPN information under one heading. If you can't find what you need in the other sections, it may be easier to use the search bar instead of digging around in Surfschool.

The live support experience

To reach live support, scroll to the bottom of the main support page, then click "Chat with us." A chat window will open in the corner of your screen. I used the live chat support to ask about receiving a link to the test account I used to write this review. As is now standard, I had to work my way past a chatbot first, but that didn't take too long. Once I reached a real person, they resolved my problem right away.

Sam Chapman for Engadget

The email ticket system is for difficult problems that can't be resolved in a single exchange, so I decided to test it with something a little more complicated. I asked about the status of IKEv2, which Surfshark promised to eliminate in 2022 but has only removed from Windows (this is in fact what they promised to do, but I wasn't aware of that yet).

I heard back about 22 hours later with a speedy and applicable answer: a representative said that "currently, Surfshark does not have plans to remove IKEv2 from all operating systems." See the next section for details about that risk.

Surfshark was founded in 2018, making it a relative newcomer to the security scene, and the VPN was its first product. Alert and Search debuted in 2019, while Antivirus and Incogni joined them in 2021. For the first two years of its existence, Surfshark was part of the Tesonet group, a Lithuanian startup incubator that also helped launch NordVPN (see "Connection to NordVPN" below).

Its youth may be why there are few serious blemishes on Surfshark's record. Like Proton VPN, which is just a year older, it simply hasn't had time to put a foot wrong yet. However, that doesn't mean it's inevitable that Surfshark will screw up at some point. So far, it appears to be learning from its predecessors and avoiding their mistakes.

The trusted root certification authority risk

The only potential error I can point to hasn't compromised anyone yet, but may in the future. In 2022, Surfshark was one of several VPNs cited for relying on a trusted root certification authority — a verification method that, if compromised, could let an attacker sneak malware onto a user's device using Surfshark's ID badge. There's no record of the certificate actually being exploited, and Surfshark claimed it was necessary to enable the IKEv2 protocol.

In response to the report, Surfshark deprecated IKEv2 from its Windows app, which was the only one that used trusted root CAs for IKEv2. It's still available on all OSes except Windows, but a Surfshark rep told me in an email that they don't use trusted root CAs on those platforms, so there's no risk.

Connection to NordVPN

In 2022, Surfshark announced a merger with NordVPN, though both companies repeatedly stressed that nothing about either service would change. According to Surfshark's announcement at the time, "The idea behind the deal is to align on a tactical level in reaching mutual goals while keeping the autonomy of our operations."

As of 2025, nothing seems especially merged about NordVPN and Surfshark. Publicly, they're still completely different brands with few mentions of the other on their respective websites. In an email, a Surfshark representative told me that "Surfshark and Nord Security operate as autonomous companies relying on separate infrastructures, different product development plans and separate customer bases."

That said, it's worth noting that Surfshark and Nord have a shared lineage in Tesonet, the Lithuania-based tech incubator that, with Nord, shares Tom Okman and Eimantas Sabaliauskas as co-founders, and nurtured Surfshark for the first two years of its life. In the end, a Surfshark representative told Engadget that Surfshark currently "operates independently" of Tesonet. Be that as it may, the combination of Surfshark and NordVPN is yet another example an increasingly consolidated VPN industry, where competitors like Kape Technologies and Ziff Davis operate multiple digital service brands under one corporate roof.

In a lot of ways, Surfshark works hard to earn your trust. Its reliance on brick-and-mortar servers testifies to that, as do its speeds, regular privacy audits and refusal to rest on its laurels. Nexus and Everlink may not be perfectly implicated, but they do have tangible results that indicate they're a lot more than just marketing gimmicks. Both are positive signs of a VPN provider genuinely trying to stand out in a crowded field.

It's also a great sign that Surfshark responded to the trusted root CA risk with a targeted intervention. I wish they'd found a way to run IKEv2 on Windows without as much risk (other VPNs manage it), but I can't deny they seem serious about addressing potential issues before they become crises. It's nice that I can enthusiastically recommend Surfshark for its attention to security as well as its speeds and straightforward features.

This article originally appeared on Engadget at https://www.engadget.com/cybersecurity/vpn/surfshark-vpn-review-a-fast-vpn-for-casual-users-170022675.html?src=rss

Apple repeats its long held criticism of the EU's Digital Markets Act

Apple has long opposed the Digital Markets Act, which is pretty much expected for a Big Tech company. Now, a bit over a year after it came into force, Apple has asked the European Commission to repeal it, according to the Financial Times and Bloomberg. "The DMA should be repealed while a more appropriate fit for purpose legislative instrument is put in place," the company has told the commission in its first public consultation to review the legislation. The EU's Digital Markets Act, which was signed into law in 2022, aims to rein in Big Tech's power and to improve competition for smaller players. If a company is found to have violated DMA's rules, it could pay between 10 to 20 percent of its global revenue. 

In a blog post separate to the feedback it submitted to the Commission, Apple detailed how DMA affects its users in the EU. The company said that because the law requires it to make sure certain features work on non-Apple products, feature rollouts get delayed in the region. Its team has not found a way to securely bring iPhone Mirroring to non-Apple devices, for instance, because it requires a lot of engineering work. As a result, Apple has yet to release that feature, along with Live Translation via AirPods, as well as Visited Places and Preferred Routes on Maps, in the European Union. 

Apple also said that since the DMA requires it to allow sideloading, other app marketplaces and alternative payment systems, users in the region are exposed to more risks. They're more exposed to things like fake banking apps and disguised malware, and they could also come across third-party payment systems that wouldn't allow refunds. If you'll recall, the European Commission slapped Apple with a $587 million fine in April for preventing developers from informing customers about sales and other offers outside the App Store. Apple called the penalty "unprecedented" and filed an appeal. 

"Regulators claimed the DMA would promote competition and give European consumers more choices," Apple wrote in its post. "But the law is not living up to those promises... That’s why we’re urging regulators to take a closer look at how the law is affecting the EU citizens who use Apple products every day." Even though the company is clearly against the DMA, it said it's dedicating "thousands of hours" to bring features to the EU. A spokesperson for the Commission told the Times that it's normal for companies to "need more time to make their products compliant" and that the Commission is helping them get there. "[C]ompliance is not optional, it’s an obligation," the spokesperson added. 

This article originally appeared on Engadget at https://www.engadget.com/big-tech/apple-repeats-its-long-held-criticism-of-the-eus-digital-markets-act-130058440.html?src=rss

One of our favorite slim MagSafe power banks is on sale for a record-low price

Carrying around charging cables, adapters or even a bulky power bank defeats the purpose of traveling light. But now there are plenty of options for those who want a power bank as svelte as their phone — even those who are investing in an iPhone Air. One of Anker's latest fits the bill: the Anker Nano 5K MagGo Slim power bank.

Now, both Anker and Amazon are running sales on it, dropping the price from $55 to $46. The 16 percent discount a new low for the power bank and available in the black and white models. It's just about a third of an inch thick and attaches right to your iPhone. On that note, it works with any MagSafe compatible phone with a magnetic case. 

Anker's Nano 5K MagGo Slim is our pick for best, well, slim MagSafe power bank. It took two and a half hours to charge an iPhone 15 from 5 percent to 90 percent. However, it could boost the battery to 40 percent in just under an hour. Overall, though, the minimalist design and easy to grip matte texture, really sold it to us.

Follow @EngadgetDeals on X for the latest tech deals and buying advice.

This article originally appeared on Engadget at https://www.engadget.com/deals/one-of-our-favorite-slim-magsafe-power-banks-is-on-sale-for-a-record-low-price-121512048.html?src=rss

Meta rolls out teen accounts for Facebook and Messenger across the world

Meta is starting to move more teens on Facebook and Messenger into dedicated "teen accounts" that have added parental controls and other protections for younger users. The expansion comes as the company says that "hundreds of millions" of teens are already using the accounts across Instagram, Facebook and Messenger.

Meta first brought teen accounts to Instagram a year ago and began rolling them out to  teens in the US, Canada, UK and Australia on Facebook and Messenger earlier this year. Now, the specialized accounts will be available to teens globally. The company has made the accounts mandatory for all teens, and requires younger teens (13 -15-year-olds) to get their parent’s permission to change safety-related settings. Meta uses AI to detect teens that may be lying about their age. The accounts allow parents to supervise how their children use Meta's apps, including features for monitoring screen time and the ability to view who their kids are messaging with. Teen accounts also come with more restrictive privacy and safety settings meant to limit their contact with adults they don't know.

Instagram is also expanding its program that allows US middle schools and high schools to expedite reports of bullying and other problematic behavior. Up to now, the company has been piloting the "school partnership program," with a handful of middle schools and high schools. Meta says that it's "heard positive feedback from participating schools" and that any US-based school can sign up to join.

The social media company has spent the last few years ramping up parental control features and attempting to close some of the more obvious gaps in its safety features. The company is currently facing numerous lawsuits and investigations into its track record on child safety.

Update, September 25, 2025, 9:03 AM ET: Clarified the age range for teen accounts.

This article originally appeared on Engadget at https://www.engadget.com/social-media/meta-rolls-out-teen-accounts-for-facebook-and-messenger-across-the-world-120000352.html?src=rss

Xreal's One Pro intrigues me in ways Meta's smart glasses don't

There I was, sitting on the couch in a nondescript suburban cafe, typing away on a MacBook Pro while staring off into the distance with what looked to be oversized glasses (which also were conspicuously plugged into the computer). To most people, I probably looked a bit silly. Nobody actually commented on my setup, or maybe I was just too focused to notice. From my perspective, I was looking at a massive 222-inch virtual display through Xreal's One Pro smart glasses, which was mirroring my laptop screen. 

Like a visual version of noise canceling, the glasses blotted out the chaos of the cafe so that I could actually concentrate on writing this review. I could also darken the frames a bit to serve as pseudo-sunglasses, further removing distractions. This work session in particular made the idea of smart glasses seem more compelling than anything in Meta's failure-prone demo for the Ray-Ban Display.

With the One Pro, Xreal is close to making the dream of versatile smart glasses a reality. It's not trying to be something you wear all day, an aspiration Meta is desperately chasing — instead, it's meant for specific purposes. Plug the One Pro into your laptop and you can use it as a virtual monitor nobody else can see. Connect it to your phone or tablet and you've got your very own personal theater wherever you go. You can even hook it up to some portable game consoles (but not the Switch 2, unfortunately) to play on the move. 

Devindra Hardawar for Engadget

Sure, the Xreal One Pro still looks clunky, with its oversized frames, thick arms and annoying USB-C cable. But at $650, it's also significantly less expensive than the Apple Vision Pro and easier to travel with than any VR headset. And at this point, it's vastly more useful than the Ray-Ban Display, since its displays work across both of your eyes and can easily mimic a full-sized TV or monitor. Meta's smart glasses can only show you a very limited amount of information in its single screen.

To be clear, that lines up with the different use case for each product: The Xreal One Pro is more of a monitor you wear on your face while stationary, whereas the Ran-Ban Display glasses aim to overlay your real-time point of view with basic phone notifications and services. Those frames are also meant to be more independent, allowing you to listen to music, take calls and capture photos and videos at a whim. But after seeing Mark Zuckerberg and his team try (and mostly fail) to demonstrate the Ray-Ban Display at its Connect developer conference, I also think the camera-less Xreal One Pro avoids ethical dilemmas inherent in most smart glasses.

Should we normalize wearable devices that can covertly record us at all times? And on a practical level, do we really want to walk around with virtual notifications in our faces all day? Personally, I think we're better off with specialized tools that leave us in control, and don't upend our entire social structure. 

Devindra Hardawar for Engadget

At the most basic level, Xreal's One Pro smart glasses are just virtual displays you can wear anywhere. You can unlock more augmented reality capabilities with the $99 Xreal Eye camera accessory, but it’s not exactly necessary. The One Pro features dual 0.55-inch Sony Micro-OLED screens running at 1080p with a 120Hz refresh rate and a 57-degree field of view (FOV), the widest we've seen yet from Xreal. Modern VR headsets like the Quest 3 typically offer a much more expansive 110-degree FOV, but they also trap you in a dark void. 

The One Pro, and all of Xreal's frames, still let you see the real world outside of its AR screens, and they don't block out ambient light. But there are also several shades you can use to darken the glasses themselves, which make the virtual display appear brighter and more distinct. At the darkest level, the One Pro can appear almost entirely black, something that's great for movies. There’s also a helpful auto transparency mode that undims the displays when you look away from your virtual screen.

Devindra Hardawar for Engadget

Bose mini-speakers are built into the arms of the glasses, as well as buttons for managing brightness and settings. A single USB-C port sits at the end of the left arm. I figured the cable would be a pain during extended sessions, but most of the time I quickly forgot it was there. Consequently, there's no wireless support — something I'm totally fine with, since that would inevitably require batteries and additional weight on the glasses. At 87 grams, the One Pro weighs about the same as a deck of playing guards. It sat on my face comfortably, thanks to its sturdy nose pads, and I was able to wear it for hours without much issue.

Since the Xreal One Pro is fashioned after sunglasses, they can't sit atop normal eyewear like the Quest 3 and other VR headsets. If you have a glasses prescription, you'll need to buy inserts from HonsVR, which start at $50. They're a bit annoying to install, but otherwise they did a fine job of letting me see the One Pro's displays clearly. They do make sharing the glasses annoying, though, since the inserts will need to be removed every time you do so. I'd also be worried about something getting bent or broken with constant removals.

I mostly used the Xreal One Pro as a virtual display while working on laptops, but I also found them incredibly helpful for watching movies and videos when away from home. On the many occasions where I was stuck waiting for my family in the hellscape of a suburban parking lot, all I had to do was put on the Xreal One Pro, plug it into my phone and I could be watching anything on a massive virtual screen. 

It was even more helpful while traveling. I could never muster the bravery to wear an Apple Vision Pro on a flight, or justify stuffing it into a bookbag. But that wasn’t a problem at all with the Xreal One Pro, as its bulbous traveling case can fit almost anywhere. Having a portable virtual display you can deploy instantly honestly feels like a superpower.

Devindra Hardawar for Engadget

As someone who's picky about displays, I was surprised how bold and colorful the glasses were when I cranked up the brightness and background shade settings. I'd definitely love to see what sharper 4K screens could look like, but even at 1080p per eye, the Xreal One Pro delivered decently clear text and detailed imagery. Its built-in Bose speakers were also surprisingly clear while watching videos or playing some background tunes. (I would always use my AirPods Pro if I really wanted to immerse myself in the movies I was watching, though.)

In a pinch, the Xreal One Pro were helpful gaming accessories too. At home, I’d rather be staring at my 4K Alienware computer monitor, or the Steam Deck OLED’s native screen. But for slower-paced titles, it was nice to game on a large virtual screen just by plugging the glasses into my Steam Deck. 

Getting the device connected to the original Nintendo Switch takes some work though, as it only works when connected to a dock and using an accessory like the Xreal Hub. Unfortunately, the Switch 2 doesn’t work with the One Pro at all right now (unless you start daisy chaining USB-C and HDMI adapters). Xreal says its upcoming Neo accessory will offer video pass-through for the Switch 2, but there’s no word on when that will arrive.

The more I used the Xreal One Pro, the more impressed I was by its sheer versatility. Its only major downsides are its $650 price (up from $600 originally), as well as the fact that you’ll never look cool wearing it. And no, you won’t be walking around while wearing the One Pro, like you could with Meta’s Ray-Ban Display. But hey, that also means you won’t have to worry about people calling you a pervert for wearing spy glasses.

This article originally appeared on Engadget at https://www.engadget.com/ar-vr/xreals-one-pro-intrigues-me-in-ways-metas-smart-glasses-dont-120000554.html?src=rss

Proton Mail's redesigned mobile app is built for speed

If you use Proton Mail on your phone, things are about to pick up. The company is rolling out new apps for Android and iOS. The updated mobile applications are rebuilt from the ground up with a "cleaner, faster and more private experience." Proton first announced in April that it was working on the revamp.

The company says the new Proton Mail lets you scroll, archive and reply twice as fast as before. It also now supports an offline mode, allowing you to read, write and organize messages while away from the internet. A redesigned interface aims for simpler navigation, with areas like the composer button now sitting within easier reach.

The iOS and Android apps, while still native to their respective platforms, now share a common codebase. Proton says they share 80 percent of their code. This should enable faster development and near-simultaneous future updates.

Proton

Product lead Anant Vijay Singh credited the update to Proton's community and business model. "The new Proton Mail mobile apps reflect this feedback and show what is possible if you build an email app without the constraints imposed by trying to monetize user data, allowing for a cleaner, faster, and more private experience," he said.

Proton has had a full plate lately. The company is working on an upcoming overhaul of Proton Calendar with similar user experience upgrades. This summer, it even joined the chatbot fracas with Lumo, which it believes can carve a niche as a more ethical AI assistant.

The Proton Mail updates begin rolling out today in the App Store and Play Store.

This article originally appeared on Engadget at https://www.engadget.com/mobile/smartphones/proton-mails-redesigned-mobile-app-is-built-for-speed-100042054.html?src=rss

The best instant cameras for 2025

Instant cameras are popular for one simple reason: they’re fun. You can bring one to a party or event and capture a spontaneous moment, then after a couple of minutes, everyone can enjoy a physical photo instead of staring at a phone screen. The quality isn’t perfect, but those snaps have a nostalgic quality that people love.

Buying one can be tricky though — you need to consider factors like price, film size, photo quality and more. Plus, there are a lot of models to choose from. This guide will help you choose the best Instant camera for your needs so you can wring the most fun possible out of it.

The main factor for most people is the size of the film. When most people think of retro instant cameras, they think of a square image size. That’s why Instagram, which used to have a Polaroid-like app icon, only supported square images for years. If that’s what you want, then Fuji’s Square or either of Polaroid’s formats are best.

However, if you want as large an image as possible, Fujifilm’s Instax wide is the way to go. For those who prefer a smaller size to pin on their fridge or slip in an envelope (or the cheaper film), Fujifilim’s Instax Mini or Polaroid’s Go format are best.

With that in mind, here are the choices. Fujifilm’s Instax alone offers three formats: Mini, Square and Wide, with film sizes of 3.4 x 2.1 inches, 3.4 x 2.8 inches and 3.4 x 4.3 inches. The corresponding image sizes are 2.44 x 1.81 inches, 2.44 x 2.44 inches and 2.44 x 3.9 inches.

Leica’s Sofort 2 and Lomography models also uses Fujifilm’s Mini format, while Polaroid’s smaller Go format is 2.64 x 2.13 inches (1.81 x 1.83-inch image) and its i-Type size is 3.46 x 4.21 inches, with a square image size of 3.11 x 3.11 inches.

As for prices, Fujifilm’s Instax films are generally around $1 - $1.20 per shot for square or wide color film and $.60 - $.75 for mini color film. Those prices go up a bit if you choose custom models with colorful or patterned borders, and down if you buy in bulk. Polaroid’s i-Type film is a bit more at about $2.00 per shot or $1.00 - $1.35 per shot for Go color film.

Image quality isn’t really the point with instant cameras of course, as they’re always going to be fuzzy compared to digital photos. However, some models (notably Fujifilm and Polaroid models) are a bit better than others.

And in terms of pricing for the cameras themselves, models with larger film are generally more expensive. It also increases if you add hybrid features like USB-C connectivity, smartphone sharing, the ability to print out photos and more, along with niceties like a selfie mirror. The best, most feature-laden cameras can cost upwards of $200 and basic versions can be well under $100.

This article originally appeared on Engadget at https://www.engadget.com/cameras/best-instant-camera-120034778.html?src=rss

Qualcomm says its new Snapdragon chips are 'the fastest and most efficient' for Windows PCs

Qualcomm has unveiled its new Snapdragon X Series chips for laptops, in addition to its new system-on-a-chip for flagship phones. The Snapdragon X2 Elite Extreme chip meant for "ultra-premium" Windows 11 laptops are designed to handle "complex, expert-level workloads" and to enable fast AI processing, as well as a multi-day battery life. Meanwhile, the Snapdragon X2 Elite chips come in two variants, one with 18 total cores and one with 12. Qualcomm claims that these processors are the "fastest, most powerful and efficient processors for Windows PCs."

The company launched the first Snapdragon X Elite chip in 2023 as its successor to the Snapdragon 8cx Gen 3 line of laptop processors. Qualcomm changed its name back then to reflect the huge leap in performance. Microsoft released a lineup of Copilot+ PCs with the new Snapdragon chip in 2024. And then earlier this year, at CES, Qualcomm unveiled the Snapdragon X for Copilot+ PCs priced $600 and under. 

Qualcomm says that the new Elite Extreme chip can power laptops for scientists and professional creators who handle "computationally intense data analytics, professional media editing and scientific research." It features the company's third-gen Oryon CPU that can apparently run at up to 75 percent faster than competitors. The chip also comes equipped with Qualcomm's Hexagon NPU, which it says is the "fastest NPU for laptops," to enable simultaneous AI tools and experiences on Copilot+ PCs.

The lower tier Elite chips can still run at up to 31 percent faster and use 43 percent less energy than the previous generation. While they're obviously meant for users who don't handle more resource-intensive workloads, Qualcomm says they also enable simultaneous AI experiences. The first laptops powered by the Snapdragon X2 Elite processors will be available in the first half of 2026. 

This article originally appeared on Engadget at https://www.engadget.com/computing/qualcomm-says-its-new-snapdragon-chips-are-the-fastest-and-most-efficient-for-windows-pcs-084535090.html?src=rss

The best streaming deals: Save on YouTube TV, Starz, Disney+ and more

Whether you’re a true cord-cutter or you just want to watch the next season of Stranger Things when it drops, everyone’s on the lookout for streaming deals nowadays. Plenty have chosen VOD and live TV streaming services over traditional cable in recent years, but the savings that choice got you just a few years ago have somewhat evaporated now. Companies like Netflix, Disney, Max and others have been consistently raising prices to the point where you may question if streaming is even worth it anymore.

We at Engadget still think so, for many reasons, but you can (and should) be smart with your money at the same time. Streaming deals are an option, even if they don’t come around with the same regularity as discounts on AirPods do. If you’re looking to save money and still stream all of the content you want, Engadget can help by laying out the best streaming deals you can get right now, how you can save with bundles and everything you should know before paying for yet another streaming service.

True streaming deals can be hard to come by. Most often, they’ll pop up during the Black Friday shopping period. On occasion, we’ll see them sparingly throughout the year and they usually take the form of a discounted monthly or annual rate for a limited period of time. Also, true streaming deals are typically on the ad-supported versions of a service, but once in a while you’ll find a unicorn of a deal on a tier that has ad-free viewing.

If you’re able to wait for a deal before subscribing to a streaming service, we recommend doing so. You’ll save money upfront and in the long run, and you also have the option to cancel your subscription before the price goes back up to the normal rate. Maybe you find you like the service so much that you’re fine paying full price for it — that’s the ideal situation. But if you’re not compelled to keep that app on rotation in your smart TV, most streaming services make it easy for you to cancel at any time. With that said, these are the best streaming deals you can snag right now.

Starz (one year) for $24 (66 percent off): Pay upfront for one year and you can get 66 percent off a Stars annual subscription. There's a month-to-month option too, which costs $4 per month for the first three months. Either option gives you access to the entire Starz TV and movie library with offline viewing and no ads.

Spotify Premium Individual (1 month) for $0 ($12 off): This is our favorite music streaming service for podcasts and social features. Right now, users who have not signed up for Spotify's Premium service before are eligible to get one month for free. The Premium Individual plan lets you listen ad-free and skip songs at will. You can also organize your listening queue and download content for offline listening. Just be aware, your subscription will auto-renew at the end of the trial period. So if you don't want to be on the hook for the $12 monthly fee, set a reminder to cancel and go back to the free version.

YouTube TV (three months) for $150 ($99 off): You can get three months of our favorite live TV streaming service for $50 per month. That should give you a decent chunk of time to see if the service is right for you while saving some cash. The discount and trial are only open to new subscribers to YouTube TV’s base plan, which includes access to over 100 channels, unlimited DVR space and six household accounts with the ability to stream on three devices at once.

Sling Orange for $23/month for the first month (50 percent off): New customers can get Sling Orange or Sling Blue for half off the usual price for the first month, bringing the final prices to $23/month and $25.50/month, respectively. Orange is likely best for sports fans, with eight exclusive sports and family channels, while Blue includes 19 exclusive news and entertainment channels. You can get both Orange and Blue access also for half off for one month, or $33 total.

DirecTV starting at $50/month for one month ($35 off): All of DirecTV's signature packages are $35 off right now for your first month when you sign up. If you opt for the base "Entertainment" package, you'll spend $50 for the first month and get access to over 90 channels, including many local stations as well as ESPN, ESPN 2 and Fox Sports 1. You'll also be able to watch on the go with the DirecTV mobile app.

Fubo Pro for $55/month for the first month ($30 off): Fubo has introductory discounts on most of its packages, and the Pro package is the least expensive plan currently listed. It offers access to 224 channels, unlimited cloud DVR and up to 10 simultaneous streams. It even includes regional sports content from the NHL, MLB and NBA.

DashPass Annual + HBO Max (with ads) for $96/year ($144 off): This offer includes access to HBO Max with ads for no extra cost when you sign up for a DashPass Annual plan. You can then decide to upgrade to Max Standard, which removes ads, for a discounted rate of $11 monthly if you want. Aside from the obvious streaming benefits, this deal gives you $0 deliver fees and lower service fees on some restaurant DoorDash orders, five percent DoorDash credits on pickup orders, on-demand grocery delivery and other members-only exclusives.

Peacock first responders discount — one year for $48 (50 percent off): Medical professionals and first responders can save 50 percent each year of Peacock. The deal requires annual verification and is open to those who work for either private or public institutions. Peacock has some great stuff to watch, including Poker Face and Killing It and more.

Student discounts on streaming services

HBO Max student discount — subscribe for $5/month (50 percent off): HBO Max offers their ad-supported tier to students for half off the usual rate. You’ll just have to verify that you’re a student through Unidays, and make note that this offer is only good for up to 12 months of service.

Hulu student discount — subscribe for $2/month (75 percent off): Those with a valid student ID can get Hulu’s ad-supported tier for 75 percent off the typical rate. They’ll keep the same sale price for as long as they’re a student as well.

Spotify student discount — Premium + Hulu with ads for $6/month (72 percent off): Spotify’s student offer continues to be one of the best around, giving you access to the Premium tier of the music streamer and Hulu’s ad-supported plan for only $6 monthly. Purchased separately, you’d pay $22 per month for both of the services. Plus, the first month is free when you sign up.

NBA League Pass student discount — one year for $120 (40 percent off): Students can get one year of League Pass for only $10 per month, which includes access to NBA TV and the ability to watch classic and archive games on-demand. On the NBA League Pass website, look for the student discount banner at the top and follow the instructions to verify your student status.

There’s more consolidation happening now than ever before in the streaming space, and that means there are more streaming bundle options. These bundles offer you access to more content with one subscription price, but those prices are typically higher than paying for a single service by itself (obviously). It may be tempting to just get the bundle, but if only one of those services in the bundle speaks to you, you’ll spend less overall by just paying for the single service.

Speaking of a deep love for a single streaming service: if all of your favorite shows are on Peacock or the latest releases on HBO Max consistently bring you joy, consider paying for one year upfront. Subscribing with an annual plan usually saves you money in the long term over paying on a monthly basis. Unfortunately, not all streaming services (looking at you, Netflix) have an annual subscription option. Here are some of the best streaming bundles you can get right now.

Disney+, Hulu, HBO Max bundle with ads for $17/month: Ad-supported HBO Max is included here, along with full, ad-supported access to Disney+ and Hulu. You’ll save 43 percent with this bundle, as opposed to paying for all three services individually.

Disney+ and Hulu Bundle Premium for $20/month: Disney and Hulu offer a few different bundles, which you can view in the drop-down lists under Choose Your Plan. This bundle removes the ads from both Disney+ and Hulu (with the exception of select live and linear content) and allows you to download content for offline viewing. You’ll save 42 percent with this bundle, as opposed to paying for both ad-free tiers individually.

Hulu + Live TV with Disney+ and ESPN+ for $96/month: This streaming bundle amalgamation is a bit confusing but it does offer a lot: you get live TV streaming via Hulu’s service plus access to the following VOD services: Hulu, Disney+ and ESPN+. Out of those three, only ESPN+ will have ads.

Disney+, Hulu, ESPN+ Bundle Basic for $17/month: You get full access to Disney+, Hulu and ESPN+ content with this package, albeit with ads across the board. This bundle price is 46 percent off the total price of all three separate subscriptions.

Disney+, Hulu, ESPN+ Bundle Premium for $27/month: Similarly to the Duo bundles, the Premium version of the Trio removes ads from most content in Disney+, Hulu and ESPN+, and you can download content for offline viewing. This price represents a 43-percent savings when compared to paying for all three ad-free tiers separately.

Sling TV + HBO Max starting at $53/month: Sling TV and HBO Max have partnered on a discount that gives new subscribers 50 percent off their first month of Sling TV, plus $5 off monthly when you subscribe to the Sling TV + HBO Max bundle. The standard price for the Sling Blue + HBO Max duo is roughly $58/month, so you'll get a monthly discount of $5 off that. In addition, for the first month only, you'll get half off the price of the bundle. The promotion also applies to the Sling Orange & Blue + HBO Max package, which has a standard price of $73/month.

Paramount+ with Showtime for $13/month or $120/year: This includes everything in Paramount+’s Essential plan, except the ads, and also provides access to Showtime content, live CBS streams and download features.

Read more streaming coverage

• The best live TV streaming services to cut cable

• The best streaming services: Netflix, Hulu, HBO Max and more

• The best streaming devices

Follow @EngadgetDeals on X for the latest tech deals and buying advice.

This article originally appeared on Engadget at https://www.engadget.com/deals/best-streaming-service-deals-133028980.html?src=rss

The Odyssey Ahead: B2C Marketers, Prepare Now For Agentic AI

Much like the start of Odysseus’ voyage after the Trojan War, marketers’ AI journey is far from over — it’s just begun. Just as marketers started experimenting with new genAI features when they emerged, agentic AI has arrived and signals the next round of transformative marketing capabilities. Tech giants are racing ahead, pouring significant resources […]

Forza Horizon 6 announced, coming to PC and Xbox in 2026

Microsoft has announced Forza Horizon 6. The game was unveiled during the Xbox Tokyo Game Show 2025 Broadcast, which makes sense, because FH 6 is set in Japan. This isn't surprising, a […]

Thank you for being a Ghacks reader. The post Forza Horizon 6 announced, coming to PC and Xbox in 2026 appeared first on gHacks Technology News.

DuckDuckGo Browser now lets you choose whether to use Search or AI on its homepage

DuckDuckGo Browser has a new option on its homepage that allows you to set AI mode as the default option. This is customizable. So, what does it do? It's a toggle that […]

Thank you for being a Ghacks reader. The post DuckDuckGo Browser now lets you choose whether to use Search or AI on its homepage appeared first on gHacks Technology News.

Ray-Ban Meta Gen 2 vs Gen 1: Which Smart Glasses Should You Buy?

There are a couple major differences between generations that could sway your opinion.

The Comic Book Guide to the ‘Wolverine’ Game

Logan's being joined by a few friends (or rather frenemies) from the comics in his blood-splattered new PS5 game.

SpaceX Wants to Fly Its Gigantic Starship Directly Over Florida

The rocket's new flight path could pose a risk to populated areas below.

YouTube Is Going to Regret This

You'll never satisfy these guys.

If You’re Dying to Try Meta’s New Smart Glasses With a Display, Now’s Your Chance

As someone who's already tried them, I'm telling you that you should do the same.

Airports in Denmark Shut Down as Mysterious Drones Spotted

Danish officials have told NATO allies the drones are from "state actors."

When It Comes to Universal’s Halloween Horror Nights, East Beats West

Universal Studios Orlando is proving to be the place to beat this spooky season, with its offerings for the 34th annual Halloween Horror Nights events.

Turns Out the First Supermassive Black Holes Weren’t So Supermassive, Research Suggests

Astronomers saw past the blinding light of a quasar, only to find a supermassive black hole that's much smaller than theoretical predictions.

ChatGPT Pulse: OpenAI Wants to Give You a Morning Briefing (and Get a Peek at Your Other Apps)

ChatGTP's newest feature, "Pulse," promises to give user personalized and relevant updates every morning.

Forecasters Are Monitoring a Wacky Storm Situation in the Atlantic

As Hurricane Gabrielle tracks east toward Azores, Tropical Storm Humberto may tangle with another developing storm system heading toward the U.S.

Microsoft Cuts Off Access to Tech That Israel Used to Surveil Palestinians

Israel collected millions of civilian phone calls made by Palestinians.

How That ‘Witcher’ Recasting Went Down—and What to Expect From New Star Liam Hemsworth

The fourth season of the hit fantasy series, with Hemsworth stepping in for Henry Cavill, arrives October 30 on Netflix.

PlayStation’s New Portable Gaming Speakers Could Be Really Good, If They Don’t Cost a Fortune

The Pulse Elevate sound like a cash grab, but Sony actually knows how to make solid audio and gaming products.

‘It’s Moving Slower Than We Want.’ Qualcomm’s Still Trying to Figure Out PC Gaming

Laptops with Qualcomm's latest chips are either great for gaming, or "are not gaming platforms."

Amazon Will Pay Up $2.5 Billion in Settlement Over Allegedly Tricking People Into Prime Subscriptions

Amazon will pay a $1 billion civil penalty—the largest ever in an FTC rule-violation case—and provide $1.5 billion in refunds to customers.

‘The Astronaut’ Teases a Returned Space Traveler’s Worst Sci-Fi Nightmare

Kate Mara plays a NASA astronaut who realizes she may have brought an alien stowaway back with her.

South Park Takes on Betting Markets, Trump, and the FCC

FCC Chairman Brendan Carr takes quite a beating.

The Director of ‘Good Boy’ on Creating Horror From a Dog’s Point of View

'Poltergeist,' 'Peanuts,' duck quacks, and plenty of treats helped bring Ben Leonberg's haunted-house tale to the screen.

The First Trailer for ‘Greenland 2: Migration’ Suggests Maybe Surviving That Killer Comet Wasn’t the Best Outcome

Gerard Butler and Morena Baccarin return to face the post-apocalypse in the sequel to the 2020 disaster film.

Dogecoin Has Made It to Wall Street

The first memecoin ETFs have hit the stock market, legitimizing the once too-online digital currency.

How Agentic AI Accelerates Healthcare Research and Innovation

Healthcare and life sciences are entering a new phase of digital transformation, powered by the rise of agentic artificial intelligence. Unlike traditional AI tools that focus on prediction or classification, agentic AI combines decision-making and tool use, giving researchers an assistant that can automate labor-intensive tasks, parse vast data sets and accelerate discovery. Dr. Ryan Ries, chief AI and data scientist at Mission Cloud Services, a CDW Company and Amazon Web Services Premier Tier Partner, says the technology is emerging as a tool to help pharmaceutical companies speed up…

Google’s AI Co-Scientist Scores Two Wins in Biology

Hey Google! What if, instead of setting reminders or fetching restaurant reviews, you helped crack the mysteries of biology?

That playful question hints at a radical vision now being tested in labs. AI systems are being recast not as digital secretaries, but as scientific partners—co-pilots built to dream up bold, testable ideas.

The pitch sounds revolutionary. But it also makes many scientists bristle. How much true novelty can a machine conjure? Isn’t it more likely to remix the past than to uncover something genuinely new?

For months, the controversy over “AI scientists” has simmered: hype versus hope, parroting versus discovery. But two new studies offer some of the strongest evidence to date that large language models (LLMs) can generate truly novel scientific ideas, leaping to non-obvious insights that might otherwise require many years of painstaking lab work. Both studies showcase Google’s AI-powered scientific research assistant, known as the AI co-scientist.

“These early examples are unbelievable—it’s so compelling,” says Dillan Prasad, a neurosurgery researcher at Northwestern University and an outside observer who has written about the potential for AI co-scientists to supercharge hypothesis generation. “You have AI agents that are producing scientific discovery! It’s absolutely exciting.”

In one of these proof-of-concept demonstrations, a team led by Gary Peltz, a liver disease researcher at Stanford Medicine, tasked the AI assistant with finding drugs already on the market that could be repurposed to treat liver fibrosis, an organ-scarring condition with few effective therapies.

He prompted the tool to look for medicines directed at epigenetic regulators—proteins that control how genes are switched on or off without altering the underlying DNA—and the AI, after mining the biomedical literature, came back with three reasonable suggestions. Peltz added two candidates of his own, and put all five drugs through a battery of tests on lab-grown liver tissue.

Two of the AI’s picks—but none of Peltz’s—reduced fibrosis and even showed signs of promoting liver regeneration in the lab tests. Peltz, who published the findings 14 September in the journal Advanced Science, hopes the results will pave the way for a clinical trial of one standout candidate, the cancer drug vorinostat, in patients with liver fibrosis.

In the second validation study, a team led by microbiologists José Penadés and Tiago Costa at Imperial College London challenged the AI co-scientist with a thorny question about bacterial evolution. The researchers had shown in 2023 that parasitic scraps of DNA could spread within bacterial populations by hitching rides on the tails of infecting viruses. But that mechanism seemed confined to one host species. How, then, did identical bits of DNA surface in entirely different types of bacteria?

So they tasked the AI with solving the mystery. They fed the system their data, background papers, and a pointed question about what hidden mechanism might explain the jump. The AI, after “thinking” and processing for two days, proposed a handful of solutions—the leading one being that the DNA fragments could snatch viral tails not just from their own host cell but also from neighboring bacteria to complete their journey.

It was uncannily correct.

What the system could not know was that Penadés and Costa already had unpublished data hinting at exactly this mechanism. The AI had, in effect, leapt to the same conclusion that it had taken the researchers years of benchwork to devise, a convergence that astonished the Imperial team and lent credibility to the tool.

“I was really shocked,” says Penadés, who at first thought the AI had hacked into his computer and accessed additional data to arrive at the correct result. Reassured that it hadn’t, he delved into the logic the AI co-scientist used for its various hypotheses and found surprising rigor. “Even for the ones that were not correct,” Penadés says, “the thinking was extremely good.”

That sound logic prompted the Imperial team to explore one of the AI’s runner-up ideas—one in which bacteria might directly pass the DNA fragments to each another. Working with microbial geneticists in France, the group is now probing that possibility further, with promising early results. “Our preliminary data seem to be pointing toward that hypothesis [also] being correct,” says Costa.

He and Penadés published both the AI’s predictions and their experimental results in the journal Cell earlier this month.

Notably, the Imperial researchers also tried various LLMs not specifically designed for scientific reasoning. These included systems from OpenAI, Anthropic, DeepSeek, and Google’s general-purpose Gemini 2.0 model. None of those jack-of-all-trades models came up with the hypotheses that proved experimentally correct.

Vivek Natarajan from Google DeepMind, who helped develop the co-scientist platform, thinks he knows what explains that edge. He points to the system’s multi-agent design, which assigns different AI roles to generate, critique, refine, and rank hypotheses in iterative loops, all overseen by a “supervisor” that manages goals and resources. Unlike a generic LLM, it grounds ideas in external tools and literature, strategically scales up compute for deeper reasoning, and vets hypotheses through automated tournaments.

According to Natarajan, academic institutions around the world are now piloting the system, with plans to expand access—though the company’s “trusted tester program” is currently at capacity and not accepting new applications. “Clearly we see a lot of potential,” he says. “We imagine that, every time you’re going to try and solve a new problem, you’re going to use the co-scientist to come along on the journey with you.”

Google is not alone in chasing this vision. In July, computer scientist Kyle Swanson and his colleagues at Stanford University described their Virtual Lab, an LLM-based system that strings together reasoning steps across biology datasets to propose new experiments.

Rivals are moving fast, too: Biomni, another Stanford-led system, is helping to autonomously execute a wide range of research tasks in the life sciences, while the nonprofit FutureHouse is building a comparable platform. Each is vying to show that its approach can turn language models into real engines of discovery.

Many onlookers have been impressed, noting that the studies offer some of the clearest evidence yet that LLMs can generate ideas worth testing at the bench. “This is going to make our jobs much easier,” says Rodrigo Ibarra Chávez, a microbiologist at the University of Copenhagen in Denmark who studies the kind of bacterial genetic hitchhiking explored by the Imperial team.

But critics warn that an over-reliance on AI-generated hypotheses in science risks creating a closed loop that recycles old information instead of producing new discoveries.

“We need tools that augment our creativity and critical thinking, not repackage existing information using alternative language,” Kriti Gaur of the life sciences analytics firm Elucidata wrote in a white paper that evaluated the Google platform. “Until this ‘AI co-scientist’ can demonstrate original, verifiable, and meaningful insights that stand up to scientific scrutiny, it remains a powerful assistant, but certainly not a co-scientist.”

The blue section of the figure shows an experimental research pipeline that led to a discovery of DNA transfer among bacterial species. The orange section shows how AI rapidly reached the same conclusions.José R. Penadés, Juraj Gottweis, et al.

Supporters counter that the latest generation of models show glimmers of what scientists might reasonably call “intelligence.” Systems like Google’s co-scientist not only recall and synthesize vast libraries but also reason through competing possibilities, discard weaker ideas, and refine stronger ones in ways that can feel strikingly human.

“I find it very invigorating,” says Peltz. “It’s like having a conversation with someone who knows more than you.”

Still, the magic doesn’t happen automatically. Extracting valuable hypotheses requires careful prompting, iterative feedback, and a willingness to engage in a kind of dialogue with the AI, notes Swanson. It’s less like pressing a button for an answer and more like mentoring a junior colleague—asking the right questions, pushing back on shallow reasoning, and nudging the system toward sharper insights.

“For now, you still need to be a bit of an expert to get the most use out of these systems,” Swanson says. “But if you ask a well-designed question, you can get really good answers.”

Google Introduces VaultGemma: An Experimental Differentially Private LLM

VaultGemma is a 1B-parameter Gemma 2-based LLM that Google trained from scratch using differential privacy (DP) with the aim of preventing the model from memorizing and later regurgitating training data. While still a research model, VaultGemma could enable applications cases in healthcare, finance, legal, and other regulated sectors.

Baidu’s PP-OCRv5 Released on Hugging Face, Outperforming VLMs in OCR Benchmarks

Baidu has released PP-OCRv5 on Hugging Face, a new optical character recognition (OCR) model built to outperform large vision-language models (VLMs) in specialized text recognition tasks. Unlike general-purpose architectures such as Gemini 2.5 Pro, Qwen2.5-VL, or GPT-4o, which handle OCR as part of broader multimodal workflows, PP-OCRv5 is purpose-built for accuracy, efficiency, and speed.

Trump Allies To Control TikTok Under New US Deal

Donald Trump on Thursday signed an executive order laying out a proposed deal for a US version of TikTok that would see Chinese ownership reduced to 20 percent and put control in the hands of the president's allies.

Judge Endorses Anthropic's $1.5 Bn Copyright Settlement

A US judge on Thursday endorsed Anthropic's deal to pay $1.5 billion to settle a class action lawsuit over amassing a library of pirated books to train its artificial intelligence.

Amazon To Pay $2.5 Bn To Settle Prime Enrollment Case

Amazon agreed Thursday to pay $2.5 billion to settle allegations from a US regulator that it used deceptive practices to enroll consumers in Amazon Prime and made it difficult to cancel subscriptions.

A Patriotic Father and Son Journey: Building FLO Business Solutions on Discipline, Family Values, and a Fight-for-Client Ethics

FLO Business Solutions is a veteran-owned, family-driven company founded by Jimmy Moore and Brandon Haile. Offering payments, payroll, POS, marketing, and lending solutions, FLO empowers small businesses with trust, service, and tailored technology.

Israel Strikes Kill Eight In Yemen's Rebel-held Capital: Huthis

Israel struck Sanaa on Thursday, killing at least eight people and wounding more than 140, Yemen's Huthis said, a day after the rebels launched a drone attack on southern Israel.

Building Brands Through Bins: Why Sustainability Matters More Than Ever For Businesses and Consumers

By redesigning trash bins, using colour coding, clear universal signage, and strategic placement, businesses can improve waste separation and reduce the costs and effort associated with recycling.

LA Entrepreneur Files Lawsuit Against AI Security Platform, Highlights Questions Around Smart Home Systems

AI-enabled security has become one of the fastest-growing segments in the smart home market, with analysts projecting billions in annual revenue over the next decade.

Brown Financials LLC Celebrates Five Years Empowering Small Businesses

The company blends flexible financing with practical training, enabling owners to meet payroll, invest in growth, and expand into new locations.

Kasey Walsh on Advocacy, Genetic Testing, and Building a Platform to Transform Families' Stories Into Change

Winsights.life envisions a HIPAA-protected space where families own their health data, track their child's progress, and connect with others navigating similar conditions.

Madagascar Police Fire Tear Gas At Protest Over Power, Water Cuts

Madagascar police fired tear gas and rubber bullets Thursday as protests in the capital Antananarivo over repeated water and electricity outages spiralled into chaos with looting and arson.

US Sanctions On Serbian Oil Firm Will Start October 1: Vucic

US sanctions on Serbia's largest oil producer, which is majority Russian-owned, will finally come into force on October 1, Serbian President Aleksandar Vucic said Thursday.

Telecom's Silent Epidemic: AB Handshake on Tackling Fraud Without Blocking Legitimate Calls

AB Handshake, a global leader in telecom fraud prevention, has developed a unique approach to tackle these challenges. The company positions itself as a trusted ally to operators, equipping them with tools to tackle fraud head-on without disrupting legitimate traffic.

Rick Krause of Pic Rite Management and Consulting Reinforces Over 30 Years of Trusted, Client-Focused Service

Founded in 1992, the company has distinguished itself by following a path very different from its competitors, one that prioritizes integrity, transparency, and an unwavering focus on client success.

Saidin M. Hernandez PA Celebrates Over a Decade of Guiding Families in Succession and Tax Planning

Since its establishment, Saidin M. Hernandez PA has provided comprehensive estate, tax, and succession planning to families and investors with domestic and international interests. Succession planning is the cornerstone of the practice.

Inmates Beheaded In Second Ecuador Prison Massacre In Days

Clashes between drug gangs on Thursday claimed at least 17 lives in the second deadly riot in an Ecuadoran prison in days, with rampaging inmates beheading and maiming rivals, officials in the violence-wracked country said.

New Sky Energy Celebrates Over 15 Years of Innovation in Clean Energy Under CEO and Chemist Deane Little

With 11 plants in the U.S. and patents secured in Europe and Japan, New Sky Energy is expanding rapidly. Korea and other Asian markets are next, with additional applications in Africa, Australia, Canada, and South America on the horizon.

Rooted in Meaning: How Xsense Authentic Places Reconnects Communities to Their True Stories

At its core, Xsense integrates creative disciplines that are too often siloed into architecture, branding, interior design, operations, and even HR into a cohesive process.

New Videomail App View'd Launches in March 2026, Ushering in a New Era of Video Communication

The vision behind View'd is rooted in a deeply personal story. The inspiration for the platform emerged when Johnson was in recovery after his double-organ transplant. During his hospital stays, he often struggled to share his progress with his family.

Why Michael Johnson Champions Herbal Medicine as a Lifestyle, Not a Cure-all

Whether it's seeking natural remedies or alternatives to symptom-driven treatments, the demand for herbal solutions is growing into a megatrend.

Trump Urges Turkey's Erdogan To Stop Buying Russian Oil

US President Donald Trump on Thursday urged Turkish counterpart Recep Tayyip Erdogan to stop buying Russian oil over the war in Ukraine, as Erdogan sought a long-awaited deal to buy US stealth fighter jets.

Germany Vows Defence Against Russia In Drone Warfare And In Space

Germany said Thursday it would step up its defences -- from drone warfare to outer space -- at a time of growing tensions between Russia and NATO while the Ukraine war grinds on.

Germany's Bosch To Cut 13,000 Jobs In Blow To Auto Sector

German industrial giant Bosch said Thursday it would cut 13,000 jobs, mostly in its auto unit, in the latest blow for the country's ailing car sector.

UN Tells Trump Probe Launched Over His 'Sabotage' Claims

The United Nations told President Donald Trump it had launched a "thorough investigation" into what the US leader called "triple sabotage" during his visit to the organization's headquarters.

Zelensky Says Russian Officials Should End War Or Find 'Bomb Shelters'

Ukrainian President Volodymyr Zelensky warned that the Kremlin will become a target and Russian officials should check for bomb shelters if Moscow does not stop its invasion of his country, Axios reported Thursday.

France's Sarkozy: From Palace To Prison

Nicolas Sarkozy entered the Elysee Palace in 2007 boasting hyperactive energy and a vision to transform France, but lost office after just one term and the ex-president is now set to go to prison in a spectacular downfall.

What Is Big Tech’s Influence on AI Development?

So, what is big tech's influence on ai development? It is both transformative and concerning.

How to Build and Publish a Docker Image to Docker Hub

Build once, run anywhere — deploy your app with Docker and Docker Hub.

10 Useful Python One-Liners for Data Engineering

Tackle everyday data engineering tasks with just one line of Python. Fast to write, easy to read, and surprisingly useful.

Climate tech out, dual-use in at The Drop

无摘要

EV charging startup Waat raises €100m, bucking climate tech downturn

无摘要

Period tracking app Flo and Google to pay $56m in privacy case

无摘要

Luxembourg Venture Days returns this October

无摘要

UK AI data centre startup Nscale raises $1.1bn round backed by Nvidia and Nokia

无摘要

Employees learn nothing from phishing security training, and this is why

A new study reveals that success is measured in the single digits in the best-case scenario. Here's what companies should do instead.

Forget the iPad: TCL's newest tablet won't break the bank or strain your eyes

The TCL Tab Nxtpaper 5G allows users to switch from full color to e-paper mode in seconds.

AI will transform work more than replace jobs, study finds

Despite fears of an impending wave of job losses, a new study shows the impact of AI on the labor market will be more nuanced.

You can claim up to $7,500 from AT&T's $177M data breach payouts - but the deadline is soon

Don't miss your chance. If you were affected by AT&T's data breaches, you can still file a claim to get your share. Here's how.

I tested the best $20 security camera you can buy (and it works outdoors)

The Blink Mini 2 is a feature-rich security camera that costs $20. An extra $10 gets you a waterproof adapter, making the deal even sweeter.

Battered by cyberattacks, Salesforce faces a trust problem - and a potential class action lawsuit

It's been a brutal year for Salesforce customers. ZDNET research reveals the CRM giant could be doing more to secure the parts of its platform exploited in recent attacks.

These XR glasses with a 200-inch screen effectively replaced my triple monitor setup

The Xreal One Pro AR glasses offer a premium viewing experience that seamlessly blends work and entertainment.

35 million Prime customers are due class action payment from Amazon - here's how much you can get

The FTC says Amazon tricked customers into signing up for Prime subscriptions and made it difficult to cancel.

Worried about your iPhone 17 scratching? Apple says don't be

People are calling out scratches on the back of new iPhone 17 models, even ones still in the store. A teardown from iFixit explains why the problem occurs, while Apple has finally responded with an explanation.

Your Android phone just got a useful Photos upgrade - and it's a big one for editors

Previously exclusive to the Pixel 10 series, you can now edit photos with a simple AI prompt - remove glare, add clouds, and more.

Would you trust AI for financial advice? That may not be as far off as you think

Several frontier models passed the world's most prestigious financial analyst exam. These topped the list.

Hisense's 136-inch micro LED TV is $50,000 off right now - how to get the deal ASAP

Hisense's first micro LED TV was $20,000 off just weeks ago. Now, you can get even cheaper.

Austria military ditches Microsoft for open-source LibreOffice - here's why

It's not about the cost savings, either. Many government organizations are replacing Microsoft software for a more important reason.

Disney+ is raising its subscription price again - here's what it'll cost you

The price hike comes at a less-than-ideal time for the streaming service.

This HP gaming laptop just dropped to $849 at Walmart - here's why that's a steal

The HP Omen 16 is one of the best-value gaming laptops I've seen in a while, housing the latest gaming hardware in a sleek design.

The tablet that easily replaced my Kindle and iPad now has a worthy successor

The TCL Nxtpaper 11 Plus is a balanced Android tablet, providing a dependable daily experience at an affordable price.

This app will pay you $30/day to record your phone calls for AI - but is it worth it?

The Neon app offers a quick way to make money. But is your privacy worth the payment?

10+ Alexa commands that every Echo user should be using ASAP

Amazon Alexa is already a great smart assistant, but there are ways to get more out of it that you may not have considered.

How to upgrade your 'incompatible' Windows 10 PC to Windows 11 - for free today

Microsoft really doesn't want customers to upgrade older PCs, but there are workarounds for many models. Here's everything you need to know before support ends in less than three weeks.

Apple iPhone 17 Pro Max vs. Samsung Galaxy S25 Ultra: I compared both flagships, and here's the winner

Which $1,000+ flagship phone should you buy in 2025? Here's how the specs compare between Apple and Samsung.

Employees learn nothing from phishing security training, and this is why

A new study reveals that success is measured in the single digits in the best-case scenario. Here's what companies should do instead.

Forget the iPad: TCL's newest tablet won't break the bank or strain your eyes

The TCL Tab Nxtpaper 5G allows users to switch from full color to e-paper mode in seconds.

AI will transform work more than replace jobs, study finds

Despite fears of an impending wave of job losses, a new study shows the impact of AI on the labor market will be more nuanced.

You can claim up to $7,500 from AT&T's $177M data breach payouts - but the deadline is soon

Don't miss your chance. If you were affected by AT&T's data breaches, you can still file a claim to get your share. Here's how.

I tested the best $20 security camera you can buy (and it works outdoors)

The Blink Mini 2 is a feature-rich security camera that costs $20. An extra $10 gets you a waterproof adapter, making the deal even sweeter.

Battered by cyberattacks, Salesforce faces a trust problem - and a potential class action lawsuit

It's been a brutal year for Salesforce customers. ZDNET research reveals the CRM giant could be doing more to secure the parts of its platform exploited in recent attacks.

These XR glasses with a 200-inch screen effectively replaced my triple monitor setup

The Xreal One Pro AR glasses offer a premium viewing experience that seamlessly blends work and entertainment.

35 million Prime customers are due class action payment from Amazon - here's how much you can get

The FTC says Amazon tricked customers into signing up for Prime subscriptions and made it difficult to cancel.

Worried about your iPhone 17 scratching? Apple says don't be

People are calling out scratches on the back of new iPhone 17 models, even ones still in the store. A teardown from iFixit explains why the problem occurs, while Apple has finally responded with an explanation.

Your Android phone just got a useful Photos upgrade - and it's a big one for editors

Previously exclusive to the Pixel 10 series, you can now edit photos with a simple AI prompt - remove glare, add clouds, and more.

Would you trust AI for financial advice? That may not be as far off as you think

Several frontier models passed the world's most prestigious financial analyst exam. These topped the list.

Hisense's 136-inch micro LED TV is $50,000 off right now - how to get the deal ASAP

Hisense's first micro LED TV was $20,000 off just weeks ago. Now, you can get even cheaper.

Austria military ditches Microsoft for open-source LibreOffice - here's why

It's not about the cost savings, either. Many government organizations are replacing Microsoft software for a more important reason.

Disney+ is raising its subscription price again - here's what it'll cost you

The price hike comes at a less-than-ideal time for the streaming service.

This HP gaming laptop just dropped to $849 at Walmart - here's why that's a steal

The HP Omen 16 is one of the best-value gaming laptops I've seen in a while, housing the latest gaming hardware in a sleek design.

The tablet that easily replaced my Kindle and iPad now has a worthy successor

The TCL Nxtpaper 11 Plus is a balanced Android tablet, providing a dependable daily experience at an affordable price.

This app will pay you $30/day to record your phone calls for AI - but is it worth it?

The Neon app offers a quick way to make money. But is your privacy worth the payment?

10+ Alexa commands that every Echo user should be using ASAP

Amazon Alexa is already a great smart assistant, but there are ways to get more out of it that you may not have considered.

How to upgrade your 'incompatible' Windows 10 PC to Windows 11 - for free today

Microsoft really doesn't want customers to upgrade older PCs, but there are workarounds for many models. Here's everything you need to know before support ends in less than three weeks.

Apple iPhone 17 Pro Max vs. Samsung Galaxy S25 Ultra: I compared both flagships, and here's the winner

Which $1,000+ flagship phone should you buy in 2025? Here's how the specs compare between Apple and Samsung.

OpenAI Releases ChatGPT ‘Pulse’: Proactive, Personalized Daily Briefings for Pro Users

OpenAI introduced ChatGPT Pulse, a proactive experience that compiles personalized, research-backed updates each morning. In preview on mobile and limited to $200/month Pro subscribers, Pulse surfaces topical cards built from a user’s chats, explicit feedback, and opt-in connected apps (e.g., calendar/email), shifting ChatGPT from a request-driven tool to a context-aware assistant. What Pulse Actually Does […]

The post OpenAI Releases ChatGPT ‘Pulse’: Proactive, Personalized Daily Briefings for Pro Users appeared first on MarkTechPost.

OpenAI Introduces GDPval: A New Evaluation Suite that Measures AI on Real-World Economically Valuable Tasks

OpenAI introduced GDPval, a new evaluation suite designed to measure how AI models perform on real-world, economically valuable tasks across 44 occupations in nine GDP-dominant U.S. sectors. Unlike academic benchmarks, GDPval centers on authentic deliverables—presentations, spreadsheets, briefs, CAD artifacts, audio/video—graded by occupational experts through blinded pairwise comparisons. OpenAI also released a 220-task “gold” subset and […]

The post OpenAI Introduces GDPval: A New Evaluation Suite that Measures AI on Real-World Economically Valuable Tasks appeared first on MarkTechPost.

Meta FAIR Released Code World Model (CWM): A 32-Billion-Parameter Open-Weights LLM, to Advance Research on Code Generation with World Models

Meta FAIR released Code World Model (CWM), a 32-billion-parameter dense decoder-only LLM that injects world modeling into code generation by training on execution traces and long-horizon agent–environment interactions—not just static source text. What’s new: learning code by predicting execution? CWM mid-trains on two large families of observation–action trajectories: (1) Python interpreter traces that record local […]

The post Meta FAIR Released Code World Model (CWM): A 32-Billion-Parameter Open-Weights LLM, to Advance Research on Code Generation with World Models appeared first on MarkTechPost.

How to Build an End-to-End Data Science Workflow with Machine Learning, Interpretability, and Gemini AI Assistance?

In this tutorial, we walk through an advanced end-to-end data science workflow where we combine traditional machine learning with the power of Gemini. We begin by preparing and modeling the diabetes dataset, then we dive into evaluation, feature importance, and partial dependence. Along the way, we bring in Gemini as our AI data scientist to […]

The post How to Build an End-to-End Data Science Workflow with Machine Learning, Interpretability, and Gemini AI Assistance? appeared first on MarkTechPost.

AI system learns from many types of scientific information and runs experiments to discover new materials

The new “CRESt” platform could help find solutions to real-world energy problems that have plagued the materials science and engineering community for decades.

Shoplifters could soon be chased down by drones

Flock Safety, whose drones were once reserved for police departments, is now offering them for private-sector security, the company announced today, with potential customers including including businesses intent on curbing shoplifting.  Companies in the US can now place Flock’s drone docking stations on their premises. If the company has a waiver from the Federal Aviation…

The Download: growing threats to vulnerable languages, and fact-checking Trump’s medical claims

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. How AI and Wikipedia have sent vulnerable languages into a doom spiral Wikipedia is the most ambitious multilingual project after the Bible: There are editions in over 340 languages, and a further 400…

Fusion power plants don’t exist yet, but they’re making money anyway

This week, Commonwealth Fusion Systems announced it has another customer for its first commercial fusion power plant, in Virginia. Eni, one of the world’s largest oil and gas companies, signed a billion-dollar deal to buy electricity from the facility. One small detail? That reactor doesn’t exist yet. Neither does the smaller reactor Commonwealth is building…

How AI and Wikipedia have sent vulnerable languages into a doom spiral

When Kenneth Wehr started managing the Greenlandic-language version of Wikipedia four years ago, his first act was to delete almost everything. It had to go, he thought, if it had any chance of surviving. Wehr, who’s 26, isn’t from Greenland—he grew up in Germany—but he had become obsessed with the island, an autonomous Danish territory,…

Mice and AI neural networks reveal similar patterns when learning to cooperate

At a time when conflict and division dominate the headlines, a new study from UCLA finds remarkable similarities in how mice and artificial intelligence systems each develop cooperation: working together toward shared goals.

'Digital brains' that 'think' and 'feel': Why do we personify AI models, and are these metaphors actually helpful?

The press has always used metaphors and examples to simplify complex issues and make them easier to understand. With the rise of chatbots powered by artificial intelligence (AI), the tendency to humanize technology has intensified, whether through comparisons to medicine, well-known similes, or dystopian scenarios.

AI can write your college essay, but it won't sound like you

Students who plan to use ChatGPT to write their college admissions essays should think twice: Artificial intelligence tools write highly generic personal narratives, even when prompted to write from the perspective of someone with a certain race or gender.

AI or humans in customer service? Preference depends on situation, large study shows

When we shop online, a chatbot answers our questions. A virtual assistant helps us track a package. And an AI system guides us through a return of our goods.

Spotify moves to tackle AI abuse with transparency measures

Spotify on Thursday unveiled several measures to encourage artists and publishers to be more transparent about their use of artificial intelligence, as well as to limit certain abuses.

Mice and AI neural networks reveal similar patterns when learning to cooperate

At a time when conflict and division dominate the headlines, a new study from UCLA finds remarkable similarities in how mice and artificial intelligence systems each develop cooperation: working together toward shared goals.

'Digital brains' that 'think' and 'feel': Why do we personify AI models, and are these metaphors actually helpful?

The press has always used metaphors and examples to simplify complex issues and make them easier to understand. With the rise of chatbots powered by artificial intelligence (AI), the tendency to humanize technology has intensified, whether through comparisons to medicine, well-known similes, or dystopian scenarios.

AI can write your college essay, but it won't sound like you

Students who plan to use ChatGPT to write their college admissions essays should think twice: Artificial intelligence tools write highly generic personal narratives, even when prompted to write from the perspective of someone with a certain race or gender.

AI or humans in customer service? Preference depends on situation, large study shows

When we shop online, a chatbot answers our questions. A virtual assistant helps us track a package. And an AI system guides us through a return of our goods.

Spotify moves to tackle AI abuse with transparency measures

Spotify on Thursday unveiled several measures to encourage artists and publishers to be more transparent about their use of artificial intelligence, as well as to limit certain abuses.

Just Released: NVIDIA HPC SDK v25.9

The new release introduces support for CUDA 13.0, and includes updated library components, bug fixes, and performance improvements.

R²D²: Three Neural Breakthroughs Transforming Robot Learning from NVIDIA Research

While today's robots excel in controlled settings, they still struggle with the unpredictability, dexterity, and nuanced interactions required for real-world...

How to Integrate Computer Vision Pipelines with Generative AI and Reasoning

Generative AI is opening new possibilities for analyzing existing video streams. Video analytics are evolving from counting objects to turning raw video content...

How to GPU-Accelerate Model Training with CUDA-X Data Science

In previous posts on AI in manufacturing and operations, we covered the unique data challenges in the supply chain and how smart feature engineering can...

Amazon to Pay $2.5 Billion in Prime Membership Settlement

The settlement is one of the largest in the history of the Federal Trade Commission, which sued Amazon two years ago.

There Are More Robots Working in China Than the Rest of the World Combined

China has embarked on a campaign to use more robots in its factories, transforming its manufacturing industries and becoming the dominant maker.

Elon Musk’s Father, Errol Musk, Accused of Child Sexual Abuse

Errol Musk has been accused of sexually abusing five of his children and stepchildren since 1993, a Times investigation found. Family members have appealed to Elon Musk for help.

Trump Clears Way for American-Owned TikTok Valued at $14 Billion

The administration has been working for months to find non-Chinese investors for a U.S. version of the app.

Microsoft Disables Some Services to Israel’s Defense Ministry

Microsoft said it found that Israel was violating some terms of service for its products and that it does “not provide technology to facilitate mass surveillance of civilians.”

Google Asks Supreme Court to Intervene in Dispute With Fortnite Creator

The case could rewrite the rules on how businesses make money on Google’s smartphone operating system.

Elon Musk’s xAI Signs Deal to Provide Grok Chatbot to US Agencies

The billionaire’s artificial intelligence company, xAI, has secured a deal with the government that will allow federal agencies to use the chatbot Grok for a small fee.

‘Peak SF’ on a Friday Night Is a Robot Fight

The artificial intelligence boom, which has brought techies flocking back to the city, has fueled a resurgence of live — and sometimes futuristic — events.

TikTok Sale Grows Closer as Trump Signs Executive Order Legalizing It

U.S. President Donald Trump has signed an executive order that outlines how the China-based Bytedance divestiture from TikTok in the United States will play out, laying the groundwork for the app to be majority-owned by U.S. persons and leaving Bytedance with a minority stake. This move only applies to the U.S.-based version of the app.

[Read More]

ProGrade Quickly Fixes SD Card Issue With Certain Canon Cameras

Canon EOS R series camera owners with ProGrade Digital SD cards should take notice of a new update to ProGrade's free Refresh Pro software. A newly discovered issue using ProGrade SD cards in certain Canon cameras has been resolved, but photographers will need to update their cards to ensure proper performance.

[Read More]

Blend In (or Stand Out) With the Peak Design x Sitka Camo Camera Strap

Peak Design has teamed up again with the hunting and outdoor gear company Sitka for a limited-edition camouflage Slide Lite camera strap. If photographers slap camouflage on their telephoto lenses, and even wear a full-blown camo cloak, a camouflage camera strap is the next logical step for shutterbugs who want to blend in.

[Read More]

Zeiss Loxia Lenses Are Discontinued as Kenko Tokina Ceases Production

Zeiss Loxia lenses have been officially discontinued according to a public notice published by Kenko Tokina and confirmed to PetaPixel by Zeiss.

[Read More]

Xiaomi is Obsessed With Proving It’s Better Than Apple


It's typical for Android smartphone makers to compare their latest products against the Apple iPhone. However, Chinese company Xiaomi has taken the rhetoric up a notch -- or two -- with its new Xiaomi 17 series, which comprises the Xiaomi 17, Xiaomi 17 Pro, and Xiaomi 17 Pro Max.

[Read More]

Gorgeous 4.2K Video of San Francisco Showcases the iPhone 17 Pro’s Capability

A filmmaker has put the new iPhone 17 Pro to the test and created a beautiful 4.2K short video of San Francisco, capturing amazing colors and dynamic range for a smartphone.

[Read More]

After Her Photos Were Seized by Police, Sally Man Predicts ‘New Era of Culture Wars’

Controversial photographer Sally Mann has warned of “a new era of culture wars” and predicted further attacks on the arts.

[Read More]

A Brand-New Camera Company is Making a Spiritual Successor to the Lumix GM5

Lifelong photographer Côme Courteault co-founded Esquisse Camera to develop a brand-new Micro Four Thirds camera that channels the spirit of popular compact cameras of the past while delivering the performance that demanding modern photographers expect.

[Read More]

OnePlus’ Next Flagship Smartphone Ditches Hasselblad’s Cameras

OnePlus teased its newest flagship phone, the OnePlus 15, at today's Snapdragon Summit China, revealing two key pieces of information. The new handheld will be powered by the brand-new Snapdragon 8 Elite Gen 5 chipset, and it will feature a new camera system that ditches Hasselblad's engineering and participation.

[Read More]

Jury Trial Will Finally Decide Fate of Andy Warhol Photographer’s Estate

A trial will finally decide the long-contested fate of Billy Name's estate, who was Andy Warhol's photographer, including the rights to his significant archive of images.

[Read More]

Canon RF 85mm f/1.4L VCM Review: Truly a Hybrid Lens

Two things that I love are 85mm lenses and Tokyo, Japan. I was fortunate to experience both, and PetaPixel's Jeremy Gray joined me for a week in Japan to visit Canon headquarters and the Utsunomiya optical factory. We missed Jordan on this trip, but Jeremy had never before visited Japan, and we wanted him to experience all that it has to offer for the first time.

[Read More]

Photographer Captures ‘Cyberpunk’ and ‘Gotham’ in the City of London

London, one of the great metropolises of the world, has been around since the days of Ancient Rome and continues to change, expand, and shift.

[Read More]

Viltrox Spark Z3 TTL Flash Offers Cyber Mech Style and Bright Light

Viltrox has announced the Spark Z3, a new on-camera flash that combines portability with practical power in a design that nods to futuristic aesthetics. Aimed at photographers seeking a reliable yet compact lighting tool, the Spark Z3 is designed to provide straightforward illumination for everyday shooting, portraits, and travel photography.

[Read More]

The Insta360 Wave is a Beamforming Voice Recorder With AI Transcription

Insta360 has announced the Wave, a beamforming "professional-grade" voice recorder designed for large rooms and groups of people. In short, it's a very high-end speakerphone with the ability to either connect with teleconferencing services or be used standalone with onboard storage.

[Read More]

BBC Invokes Iconic War Photos in Film Calling for Western Journalists to Enter Gaza

The BBC, AFP, AP, and Reuters have collaborated on a film that uses powerful news photographs from past conflicts to demand access to Gaza.

[Read More]

Biden’s Portrait in Trump’s New Presidential Walk of Fame is a Photo of an Autopen

President Donald Trump has hung a photo of an autopen to represent President Joseph Biden in the newly designed presidential portrait gallery in the West Wing Colonnade.

[Read More]

AI is helping judges to quickly close cases, and lawyers to quickly open them

Brazilian Supreme Federal Court building in Brasilia, DF, Brazil.

WWI-Era Shipwrecks in Mallows Bay Form Ecological Sanctuary

Nearly 100 years ago dozens of ships were abandoned in a shallow bay in the Potomac River. Today plants and animals are thriving on the skeletons of these vessels

Taylor Swift’s Speech Pattern Changed over Time, Linguistics Study Shows

An analysis of Taylor Swift’s interviews suggests her speech pattern has changed over her career

Punch Cards, Pipeline Problems, and the Future of Women in Computing

Carla Brodley, founding executive director of the Center for Inclusive Computing at Northeastern University, explains how to make computer science education more accessible to everyone

Good Conversations Don’t Require Everybody to Agree, Neuroscience Shows

Brain imaging is illuminating the patterns linked to productive, positive dialogue, and those insights could help people connect with others

Bird Flu and Human Flu Viruses Could Mix in Cow Udders and Spark a Pandemic

Cells in cow udders could act as a site for human flu and bird flu viruses to swap genes and generate dangerous novel strains

Climate and healthcare projects among new recipients of €33m DTIF fund

To date, 223 SMEs have been awarded funding via the programme, which aims to positively disrupt Ireland’s most critical sectors.

Read more: Climate and healthcare projects among new recipients of €33m DTIF fund

‘Trust, support and challenges have been invaluable in shaping my career’

Workhuman’s Noel Barry discusses his role as a senior director for back office engineering.

Read more: ‘Trust, support and challenges have been invaluable in shaping my career’

UCC wants women to share their healthcare experience in Ireland

The study wants to hear from adults in Ireland who have experienced conditions such as endometriosis and sought a diagnosis in the country.

Read more: UCC wants women to share their healthcare experience in Ireland

EV charging and solar power provider ePower raises €30m

ePower’s landmark fundraise, along with other major announcements from PCRE and Ørsted, marked a significant week for Ireland’s renewable energy industry.

Read more: EV charging and solar power provider ePower raises €30m

No link between paracetamol use in pregnancy and autism, major studies show

Dr Renee Gardner and Dr Viktor H Ahlqvist from the Karolinska Institutet and Dr Brian Lee from Drexel University studied nearly 2.5m children in Sweden.

Read more: No link between paracetamol use in pregnancy and autism, major studies show

Ireland trials new emergency comms platform for disaster relief

The new Mission Critical Communications system is built by Vodafone Ireland in partnership with the Irish Government.

Read more: Ireland trials new emergency comms platform for disaster relief

Should you invest in micro-credentials or a full-time degree?

Education is no longer a one-size-fits-all operation and how you choose to learn can be pivotal to future opportunities.

Read more: Should you invest in micro-credentials or a full-time degree?

Trump expected to sign order facilitating TikTok sale

Earlier this week, Trump suggested that media mogul Rupert Murdoch and his son Lachlan could be involved in the sale.

Read more: Trump expected to sign order facilitating TikTok sale

Inpute Technologies acquires Limerick’s Treaty Digital, plans new jobs

The company plans to 'significantly expand' the headcount in the Limerick office , the Inpute CEO said.

Read more: Inpute Technologies acquires Limerick’s Treaty Digital, plans new jobs

Scrap Digital Markets Act, Apple tells EU

'DMA isn’t helping markets. It’s making it harder to do business in Europe,' Apple says.

Read more: Scrap Digital Markets Act, Apple tells EU

Improved Gemini 2.5 Flash and Flash-Lite

Improved Gemini 2.5 Flash and Flash-Lite

Two new preview models from Google - updates to their fast and inexpensive Flash and Flash Lite families:

The latest version of Gemini 2.5 Flash-Lite was trained and built based on three key themes:

• Better instruction following: The model is significantly better at following complex instructions and system prompts.
• Reduced verbosity: It now produces more concise answers, a key factor in reducing token costs and latency for high-throughput applications (see charts above).
• Stronger multimodal & translation capabilities: This update features more accurate audio transcription, better image understanding, and improved translation quality.

[...]

This latest 2.5 Flash model comes with improvements in two key areas we heard consistent feedback on:

• Better agentic tool use: We've improved how the model uses tools, leading to better performance in more complex, agentic and multi-step applications. This model shows noticeable improvements on key agentic benchmarks, including a 5% gain on SWE-Bench Verified, compared to our last release (48.9% → 54%).
• More efficient: With thinking on, the model is now significantly more cost-efficient—achieving higher quality outputs while using fewer tokens, reducing latency and cost (see charts above).

They also added two new convenience model IDs: gemini-flash-latest and gemini-flash-lite-latest, which will always resolve to the most recent model in that family.

I released llm-gemini 0.26 adding support for the new models and new aliases. I also used the response.set_resolved_model() method added in LLM 0.27 to ensure that the correct model ID would be recorded for those -latest uses.

llm install -U llm-gemini

Both of these models support optional reasoning tokens. I had them draw me pelicans riding bicycles in both thinking and non-thinking mode, using commands that looked like this:

llm -m gemini-2.5-flash-preview-09-2025 -o thinking_budget 4000 "Generate an SVG of a pelican riding a bicycle"

I then got each model to describe the image it had drawn using commands like this:

llm -a https://static.simonwillison.net/static/2025/gemini-2.5-flash-preview-09-2025-thinking.png -m gemini-2.5-flash-preview-09-2025 -o thinking_budget 2000 'Detailed single line alt text for this image'

gemini-2.5-flash-preview-09-2025-thinking

A minimalist stick figure graphic depicts a person with a white oval body and a dot head cycling a gray bicycle, carrying a large, bright yellow rectangular box resting high on their back.

gemini-2.5-flash-preview-09-2025

A simple cartoon drawing of a pelican riding a bicycle, with the text "A Pelican Riding a Bicycle" above it.

gemini-2.5-flash-lite-preview-09-2025-thinking

A quirky, simplified cartoon illustration of a white bird with a round body, black eye, and bright yellow beak, sitting astride a dark gray, two-wheeled vehicle with its peach-colored feet dangling below.

gemini-2.5-flash-lite-preview-09-2025

A minimalist, side-profile illustration of a stylized yellow chick or bird character riding a dark-wheeled vehicle on a green strip against a white background.

Artificial Analysis posted a detailed review, including these interesting notes about reasoning efficiency and speed:

• In reasoning mode, Gemini 2.5 Flash and Flash-Lite Preview 09-2025 are more token-efficient, using fewer output tokens than their predecessors to run the Artificial Analysis Intelligence Index. Gemini 2.5 Flash-Lite Preview 09-2025 uses 50% fewer output tokens than its predecessor, while Gemini 2.5 Flash Preview 09-2025 uses 24% fewer output tokens.
• Google Gemini 2.5 Flash-Lite Preview 09-2025 (Reasoning) is ~40% faster than the prior July release, delivering ~887 output tokens/s on Google AI Studio in our API endpoint performance benchmarking. This makes the new Gemini 2.5 Flash-Lite the fastest proprietary model we have benchmarked on the Artificial Analysis website

<p><small></small>Via <a href="https://news.ycombinator.com/item?id=45375845">Hacker News</a></small></p>


<p>Tags: <a href="https://simonwillison.net/tags/google">google</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/llm">llm</a>, <a href="https://simonwillison.net/tags/gemini">gemini</a>, <a href="https://simonwillison.net/tags/pelican-riding-a-bicycle">pelican-riding-a-bicycle</a>, <a href="https://simonwillison.net/tags/llm-reasoning">llm-reasoning</a>, <a href="https://simonwillison.net/tags/llm-release">llm-release</a>, <a href="https://simonwillison.net/tags/artificial-analysis">artificial-analysis</a></p>

MIT’s SCIGEN Tool Steers AI Models Toward Breakthrough Quantum Materials

Researchers at MIT have unveiled SCIGEN, a new tool that directs generative AI models to design materials with exotic quantum properties. The advance could accelerate discoveries critical for quantum computing and next-generation electronics.

Overcoming the Quantum Materials Bottleneck

Generative models from Google, Microsoft, and Meta have produced tens of millions of new material candidates, but they often fail when tasked with creating materials with rare quantum behaviors such as superconductivity.

MIT’s team addressed this challenge by developing SCIGEN — short for Structural Constraint Integration in GENerative model — to guide models toward promising crystal structures.

“The models from these large companies generate materials optimized for stability,” said Mingda Li, MIT’s Class of 1947 Career Development Professor. “Our perspective is that’s not usually how materials science advances. We don’t need 10 million new materials to change the world. We just need one really good material.”

How SCIGEN Works

SCIGEN integrates with diffusion models, a popular class of generative AI, and enforces geometric constraints during every generation step. This process steers models toward producing materials with structures known to host quantum phenomena, such as Kagome and Archimedean lattices.

When applied to DiffCSP, a leading AI materials generator, SCIGEN produced more than 10 million candidate materials with Archimedean lattices. One million passed stability screening. Using Oak Ridge National Laboratory’s supercomputers, researchers simulated 26,000 of these materials and found magnetic properties in 41 percent.

Two previously undiscovered compounds, TiPdBi and TiPbSb, were synthesized and confirmed to exhibit the predicted properties. “Archimedean lattices give rise to quantum spin liquids and so-called flat bands, which can mimic the properties of rare earths without rare earth elements, so they are extremely important,” said co-author Mouyang Cheng.

Accelerating Quantum Computing Research

Quantum spin liquids could provide the foundation for robust, error-resistant qubits. However, experimental discovery of such materials has been slow. SCIGEN could offer researchers thousands of promising candidates to test, significantly reducing the time from computational prediction to laboratory validation.

“This work presents a new tool, leveraging machine learning, that can predict which materials will have specific elements in a desired geometric pattern,” said Steve May, professor at Drexel University. “This should speed up the development of previously unexplored materials for applications in next-generation electronic, magnetic, or optical technologies.”

Looking Ahead

While SCIGEN has demonstrated success, researchers emphasize the continued importance of experimental validation. Future work could extend SCIGEN’s capabilities to include chemical and functional constraints, further refining its ability to generate materials with real-world applications.

By combining human insight with AI’s generative power, SCIGEN offers a pathway to discovering the rare, transformative materials needed to unlock the next era of quantum technologies.

Join thousands of practitioners at ODSC AI West 2025, the leading applied data science conference. Gain hands-on training in generative AI, LLMs, RAG, AI Safety, and more through expert-led workshops and bootcamps. Explore cutting-edge tools in the AI Expo Hall, connect with industry leaders, and customize your experience with flexible 1- to 4-day passes. Don’t miss this chance to expand your AI skills and network — register now to secure your spot.

CrewAI and the Future of Autonomous Agents with João Moura

Autonomous AI agents have moved from hype demos to real enterprise deployments, and few frameworks embody this shift better than CrewAI. What began as an open-source side project in late 2023 has rapidly grown into one of the most widely used agent platforms, powering hundreds of millions of automations each month. More importantly, it has become a proving ground for how organizations of every size can adopt multi-agent systems to drive measurable business outcomes.

In a recent ODSC Ai X Podcast, João “Joe” Moura, founder and CEO of CrewAI, shared his perspective on the current state of agents, what makes CrewAI unique, and why we’re still only in the early chapters of agent adoption. His insights reveal not just the trajectory of CrewAI, but also where autonomous agents fit into the broader AI landscape.

You can listen to the full podcast on Spotify, Apple, and SoundCloud.

From Side Project to Enterprise Platform

CrewAI’s origins highlight a recurring pattern in technology: personal need sparks innovation that scales far beyond its creator’s intent. Moura initially built a small agent to help him write LinkedIn posts more efficiently. The simplicity of that experiment exposed a bigger truth — the barrier to building and deploying useful agents was unnecessarily high. By abstracting away complexity, he saw an opportunity to make agent creation accessible to developers and non-technical users alike.

Released on GitHub in late 2023, CrewAI quickly gained traction. Within months, enterprise teams were experimenting with it to orchestrate agents that could collaborate on complex tasks. When companies like Oracle approached Moura asking for support, it became clear this was more than a hobby project. Today, CrewAI balances its thriving open-source ecosystem with an enterprise-grade platform designed for security, governance, and scale.

Defining What Makes an Agent

The AI community often debates what qualifies as an “agent.” Some frameworks chain prompts together and label them agents, while others stretch the definition to include any workflow powered by a large language model. Moura’s view — and CrewAI’s guiding principle — is refreshingly simple: an agent must have agency.

In practice, that means the model itself decides the flow of execution. If the application is strictly following pre-programmed if/then rules and only calling the LLM when instructed, it’s not truly an agent. By focusing on this minimal but essential definition, CrewAI keeps complexity in check while still enabling powerful orchestration. As Moura puts it, “An agent needs agency, otherwise it’s just another script”.

Early Days, Real Value

Despite explosive growth in usage — CrewAI now supports more than 475 million automations per month — the broader market is still in its “teenage years.” Most companies are in pilot or proof-of-concept mode, exploring what agents can do but not yet fully operationalizing them across the organization.

That said, value is already being realized. Some CrewAI customers have scaled from running 10,000 agent workflows per month to more than half a million. Enterprises across industries — from consulting giants like PwC to consumer brands like PepsiCo — are deploying CrewAI to tackle diverse use cases:

• Go-to-market operations: sales prospecting, marketing personalization, training materials, and customer support.
• Internal GPT alternatives: secure, domain-specific research assistants that don’t rely on public LLMs.
• Technical automation: agents that write and maintain code, update enterprise systems like Salesforce or SAP, and handle repetitive IT tasks.

The adoption curve may be early, but momentum is undeniable.

Empowering Non-Technical Users

One of CrewAI’s most impactful innovations is lowering the barrier to entry. While early adopters of agent frameworks were almost exclusively engineers, Moura argues the future depends on empowering “citizen developers” — legal teams, HR professionals, procurement specialists — who want the same agent capabilities without deep technical skills.

CrewAI’s Studio feature embodies this vision. Users can “chat their way” into a use case, describing a desired workflow in natural language. The platform translates that into an agent orchestration under the hood. For organizations, this democratization means powerful automations can emerge from business units themselves, not just centralized IT teams. Engineers still play a role, often by building shared tool repositories or adding guardrails, but their impact is multiplied when non-technical colleagues can create agents safely.

Why Some AI Projects Fail

Headlines often highlight the failures of AI adoption. A recent MIT report claimed 95% of companies investing in generative AI see little or no return. Moura sees the reality differently. The problem, he argues, is rarely the technology — it’s organizational readiness. Fear, lack of clear success criteria, and weak change management derail projects more often than model performance.

CrewAI tackles this head-on by helping enterprises define success metrics, prioritize ROI, and embed security and governance from day zero. By qualifying customers and ensuring they’re ready to scale, CrewAI avoids the trap of impressive demos that never translate into business value.

Practical Steps for Adoption

So how should organizations begin integrating agents into their workflows? Moura suggests starting small and simple:

1. Target low-complexity, low-precision tasks — such as scheduling, email batching, or calendar management. These quick wins build confidence.
2. Progress to larger automations — where ROI becomes significant, such as procurement, churn prevention, or financial reporting.
3. Embed guardrails early — governance, observability, and security shouldn’t be afterthoughts, especially as non-technical staff begin building agents.
4. Measure success continuously — companies that clearly define ROI and track outcomes are best positioned to expand adoption.

This crawl-walk-run approach helps enterprises avoid “failure to launch” and ensures agent deployments align with business goals.

The Technology Landscape

While new LLM releases and benchmarks generate excitement, Moura is clear-eyed about their practical impact. For 99% of enterprise automations, today’s models are already “good enough.” The bottleneck isn’t model capability but orchestration, governance, and cost.

He also sees standards like the Model Context Protocol (MCP) and emerging agent-to-agent communication protocols as promising, though still in flux. CrewAI’s strategy is pragmatic: focus on proven protocols, adopt standards where they add value, and avoid overengineering. Security and governance — especially around data permissions and prompt injection risks — remain top concerns for enterprise customers.

Looking Ahead: Agents as the New Operating System

Where is this all headed? Moura believes we are on the cusp of a fundamental shift in how people work. Just as operating systems shaped the last era of software, autonomous agents could serve as the next layer of abstraction — handling orchestration, anticipating needs, and integrating seamlessly across applications.

With AI already embedded into daily workflows, from email drafting to data analysis, the transition is underway. What’s different now is the level of attention and urgency: when CEOs of the world’s largest tech companies are coding on camera and championing AI adoption, the signal is clear. The genie is out of the bottle, and agents like those built on CrewAI are set to redefine the future of work.

Conclusion

CrewAI’s rise from an open-source project to an enterprise platform illustrates both the promise and the challenges of autonomous agents. By emphasizing simplicity, agency, and accessibility, it has carved out a unique space in a crowded AI ecosystem. While the market is still early, the combination of measurable value, democratized access, and enterprise-ready governance suggests that agents are not a passing trend — they’re a new foundation.

For organizations, the lesson is straightforward: don’t wait for the next model release. The tools available today are powerful enough to deliver real impact. The question is not whether to explore agents, but how quickly you can build, measure, and scale them. And as CrewAI shows, the sooner you start, the sooner you’ll unlock that value.

Agent2Agent and MCP for RAG, AI IDEs, LLM Post Training, Voice AI, and the ODSC AI West Expo Pass…

Agent2Agent and MCP for RAG, AI IDEs, LLM Post Training, Voice AI, and the ODSC AI West Expo Pass Benefits

Rethinking RAG: Building Smarter AI Agents with Agent2Agent and MCP

With Agent2Agent, each AI agent can be modular, focused on a single job, just like a DLL in Windows or a shared object in Linux. This modularity makes agents easier to build, test, and, more importantly, reuse.

The Rise of AI-Powered Integrated Development Environments

AI in software development is poised to elevate our coding environments into “super-IDEs” that automate routine work and help engineers focus on creative problem-solving.

CrewAI and the Future of Autonomous Agents with João Moura

Autonomous agents are no longer just hype. CrewAI is powering 475M+ automations monthly, helping companies scale from small pilots to enterprise-wide adoption.

Video of the Week: Introduction to LLM Post Training by Maxime Labonne, PhD

In this talk, we will cover the fundamentals of modern LLM post-training at various scales with concrete examples. We will explore key training techniques, including supervised fine-tuning and preference alignment. The presentation will delve into evaluation frameworks with their pros and cons for measuring model performance.

ODSC Highlights

10 In-Demand AI Skills You’ll Learn at ODSC AI West 2025

From mastering the basics of data science to becoming a pro with LLMOps and agentic AI, these are ten in-demand skills you’ll learn at ODSC AI West in October.

What Can You Do With a Free ODSC AI West Expo Pass?

With an Expo Hall, keynote sessions, and networking opportunities, there’s plenty you can do with a free ODSC AI West Expo+ pass.

5 Ways to Get Enterprise AI Going at ODSC AI West

Discover how to scale enterprise AI with expert-led sessions at ODSC AI West 2025 & the Gen AI X Summit this October.

AI Startup Showcase at ODSC AI West
The AI Startup Showcase at ODSC AI West 2025 is your chance to connect with the next generation of AI innovators. Founders can pitch their vision to an audience of investors, partners, and peers, while attendees can get a front-row seat to the boldest ideas shaping AI today.

Gen Ai X Summit Passes Available

Co-located with ODSC West in the heart of Silicon Valley, the Gen Ai X Summit brings together industry leaders and pioneers at the forefront of applications and innovation from across the country.

Volunteer at ODSC AI West 2025

Ready to become an integral part of the best AI conference around? Volunteer for ODSC AI West this October and get a free conference pass in exchange for your help.

Industry, Opinion, Career Advice

Inside Google’s AI Stack: From Gemini to Robotics, Video, and Voice

Explore Google’s AI stack — from Gemini reasoning models to VEO video, Lyria music, robotics, and voice APIs — reshaping AI for developers and creators.

The AI Advantage: How Creators Can Use AI to Generate Portfolio Concepts

Discover how AI can help creators beat creative block, visualize ideas, and craft standout portfolios — without replacing human creativity.

Ready to move from theory to mastery?

At ODSC AI West 2025, you’ll not only keep pace but gain hands-on skills across the fields that are changing fastest: AI Engineering, AgentOps, Evaluation, Data Science, Agentic AI, and more.

Register now for 40% off.

Data Science & AI News

Amazon Introduces AI-Powered Tools to Accelerate Product Launches for Sellers

Amazon introduced AI-powered tools to help sellers launch products faster, cut inventory risks, and improve early success with insights, reviews, and real-time coaching.

New Research Highlights Scheming Risks in AI Models — and Promising Mitigation Methods

OpenAI and Apollo Research revealed new findings on AI scheming, showing significant risk reduction using deliberative alignment while urging industry-wide transparency and proactive safeguards.

Citigroup Pilots AI Agent Capabilities to Automate Complex Workflows

Citigroup is piloting AI agent capabilities in its proprietary platform, Citi Stylus Workspaces. The pilot tests autonomous task completion across internal systems, raising questions about cost, ROI, and workforce impact.

How Developers are Using AI: Insights From the 2025 DORA Report

Discover key findings from the 2025 DORA Report on AI-assisted software development, including adoption rates, productivity gains, and trust insights.

MIT’s SCIGEN Tool Steers AI Models Toward Breakthrough Quantum Materials

MIT researchers have developed SCIGEN, a generative AI tool that guides models to create quantum materials with exotic properties, accelerating the search for breakthroughs in superconductivity and quantum computing.

New Podcast Episode: Voice AI is About to Get Loud with Kwindla Kramer

This episode explores why voice is poised to become the primary interface for AI, what makes voice agents so challenging to build, and how real-world use cases — from call centers to appointment scheduling — are proving the value of well-designed voice AI systems.

Spotify | Apple | SoundCloud

Upcoming Webinars, Meetups, and Ai+ Live Training Sessions

Building and Deploying Your First Agent with Tools on ADK

Wednesday, October 1st, 12:00 PM ET

Join us for this 1-hour webinar where you’ll learn how to build and deploy your very first AI agent with tools using the Agent Development Kit (ADK), an open-source, code-first Python toolkit from Google. We will demystify the core concepts of agents and tools, and guide you through a practical, step-by-step process to create a functional agent that can access and use external data.

ODSC & Google Cloud AI Meetup in Seattle, WA
Wednesday, October 1st, 6:00 PM — 8:00 PM PDT

Join us for an evening of connection, learning, and community as ODSC & Google Cloud bring the AI conversation to Seattle. Whether you’re deploying AI systems or just starting your journey, you’ll meet people and ideas that can spark your next project.

Webinar: Introduction to AI in Robotics

Tuesday, October 14th, 2025 04:00 PM

This session will provide a foundational understanding of how AI is revolutionizing the field of robotics, moving beyond traditional, pre-programmed systems to create intelligent, autonomous machines.

Microsoft Unveils Microfluidic Cooling Breakthrough for AI Chips

Microsoft has introduced a new microfluidic cooling system that could transform data center efficiency and sustainability. The technology cools AI chips up to three times more effectively than today’s advanced cold plate systems, a key development as next-generation GPUs generate unprecedented levels of heat.

Cooling at the Silicon Level

The innovation etches tiny microchannels — similar in size to a human hair — directly into the back of a silicon chip. Liquid coolant flows through these channels, pulling heat away at the source rather than through several insulating layers.

Microsoft researchers also used AI to analyze unique thermal patterns across the chip and optimize coolant flow, improving performance and reducing maximum temperature rise by up to 65%.

“Microfluidics would allow for more power-dense designs that will enable more features that customers care about and give better performance in a smaller amount of space,” said Judy Priest, corporate vice president and CTO of Cloud Operations and Innovation at Microsoft.

[rcblock id=”43608"]

Benefits for Datacenter Efficiency

Current cooling approaches rely on air or cold plates, which limit how tightly servers can be packed together. By cooling chips directly, microfluidics could allow Microsoft to increase server density, reduce energy used for chilling coolant, and lower datacenter operational costs. The technique also enables safe overclocking — pushing servers harder during peak workloads — without damaging chips.

“If you’re still relying heavily on traditional cold plate technology, you’re stuck,” said Sashi Majety, senior technical program manager for Microsoft’s Cloud Operations and Innovation group.

Bio-Inspired Design and AI Optimization

To perfect the system, Microsoft collaborated with Swiss startup Corintis and used AI to create bio-inspired channel designs modeled after the branching patterns in leaves and butterfly wings. These optimized layouts distribute coolant more efficiently than straight channels and minimize the risk of clogging while maintaining chip strength.

https://medium.com/media/3c0bfe037a5bedc0cca6d4f527c1b572/href

Preparing for the Future of AI Workloads

The breakthrough comes as Microsoft invests over $30 billion in infrastructure this quarter, including the development of its Cobalt and Maia chips. Microfluidics may play a role in future chip generations, including 3D-stacked architectures that would otherwise overheat under traditional cooling methods.

“Microfluidics improves cost, reliability, speed, consistency of behavior, and sustainability,” said Jim Kleewein, Microsoft technical fellow for Microsoft 365 Core Management. “We want microfluidics to become something everybody does.”

If widely adopted, microfluidic cooling could reshape datacenter design, allowing more compute power in less space and supporting the growing demand for AI services.

Join thousands of practitioners at ODSC AI West 2025, the leading applied data science conference. Gain hands-on training in generative AI, LLMs, RAG, AI Safety, and more through expert-led workshops and bootcamps. Explore cutting-edge tools in the AI Expo Hall, connect with industry leaders, and customize your experience with flexible 1- to 4-day passes. Don’t miss this chance to expand your AI skills and network — register now to secure your spot.

AI-Generated “Workslop” Threatens Workplace Productivity

Generative AI is now deeply embedded in the workplace, but its impact on productivity is raising red flags. According to a new study from BetterUp Labs in collaboration with the Stanford Social Media Lab, the rapid adoption of AI tools is generating a surge of low-quality work output — dubbed “workslop” — that creates more problems than it solves.

What Is Workslop?

Workslop refers to AI-generated work that appears polished but lacks the depth or context needed to advance a task. Instead of saving time, it shifts the burden downstream, forcing colleagues to decipher, correct, or redo the work.

This issue is widespread. In a survey of 1,150 U.S. full-time employees, 40% reported receiving work slop in the past month. Respondents estimated that 15.4% of workplace content falls into this category, with professional services and technology sectors hit the hardest.

The Hidden Productivity Tax

The costs of workslop are significant. Employees reported spending nearly two hours addressing each instance, translating into an estimated $186 per employee per month in lost productivity. For an organization with 10,000 employees, this amounts to more than $9 million annually.

The social cost may be even greater. More than half of employees said they view colleagues who send workslop as less creative, capable, or reliable. Forty-two percent said their trust declined, and a third said they are less willing to work with that colleague in the future.

Why It Happens

Generative AI makes it easy to produce slides, summaries, and reports quickly. Some employees use this capability to enhance quality, but others lean on AI to avoid doing the hard thinking themselves. The result is work that is formatted well but contextually weak, pushing effort and decision-making onto others.

As one finance professional in the study put it, “It created a situation where I had to decide whether I would rewrite it myself, make him rewrite it, or just call it good enough. It is furthering the agenda of creating a mentally lazy, slow-thinking society.”

How Leaders Can Respond

Experts caution that indiscriminate “AI-first” mandates may encourage thoughtless use of these tools. Instead, leaders should:

• Provide clear guidelines: Define when AI is appropriate and how outputs should be reviewed before sharing.
• Model intentional use: Show teams how to use AI as a creativity enhancer, not a shortcut.
• Foster collaboration: Encourage feedback and context-sharing so AI-generated work supports, rather than undermines, colleagues.

BetterUp’s researchers note that employees with high agency and optimism — “pilots” rather than “passengers” — use AI more purposefully, leading to better outcomes. Leaders can amplify this effect by promoting thoughtful AI adoption and integrating AI workflows into broader collaboration practices.

The Bottom Line

Workslop may feel effortless to produce, but it carries a steep cost in time, money, and trust. Companies that set guardrails, encourage purposeful use, and prioritize collaboration will be better positioned to turn AI adoption into measurable productivity gains rather than a costly distraction.

Join thousands of practitioners at ODSC AI West 2025, the leading applied data science conference. Gain hands-on training in generative AI, LLMs, RAG, AI Safety, and more through expert-led workshops and bootcamps. Explore cutting-edge tools in the AI Expo Hall, connect with industry leaders, and customize your experience with flexible 1- to 4-day passes. Don’t miss this chance to expand your AI skills and network — register now to secure your spot.

How Developers are Using AI: Insights From the 2025 DORA Report

Google Cloud’s 2025 DORA Report reveals how dramatically AI has reshaped software development. The annual research surveyed nearly 5,000 technology professionals worldwide, uncovering that AI has moved from a niche experiment to a standard part of the development workflow.

AI Adoption Nears Universality

TheORA Report shows AI adoption among software professionals has surged to 90%, a 14% increase from 2024. Developers, product managers, and engineers now spend a median of two hours daily working with AI tools.

A striking 65% report heavy reliance on AI in their development process, with 20% saying they rely on it “a lot” and 8% reporting “a great deal.”. “AI is no longer a novelty, but a near-universal part of a developer’s toolkit,” notes the report.

Productivity and Code Quality Gains

The 2025 DORA Report highlights clear benefits for teams integrating AI:

• Productivity: Over 80% of respondents said AI improved their efficiency.
• Code Quality: 59% reported that AI positively impacted code quality.
• Throughput: AI adoption is now linked to higher software delivery rates, reversing last year’s slowdown.

The Trust Paradox

Despite these gains, trust remains mixed. While 24% of respondents report a high level of trust in AI outputs, 30% trust AI only a little or not at all. This indicates that AI is widely used as a supportive tool rather than a full replacement for human decision-making.

AI as a “Mirror and Multiplier”

The 2025 DORA Report findings describe AI as both a mirror — exposing organizational weaknesses — and a multiplier, amplifying productivity in cohesive teams. To dig deeper, the report identifies seven team archetypes, from “Harmonious High-Achievers” to “Legacy Bottlenecks,” offering organizations a way to benchmark performance and culture.

The DORA AI Capabilities Model

Adoption alone is not enough to unlock AI’s full potential. The DORA AI Capabilities Model, introduced in this report, outlines seven technical and cultural capabilities that maximize AI’s impact. This framework gives organizations a practical blueprint to evolve workflows, processes, and team structures to support AI-driven development.

Key Takeaway

AI is transforming software development — improving productivity, code quality, and release velocity. However, success depends on more than just adoption. Organizations must foster the right culture, processes, and technical foundations to fully realize AI’s promise.

Readers can download the full State of AI-Assisted Software Development Report for deeper insights and the complete capabilities model.

Join thousands of practitioners at ODSC AI West 2025, the leading applied data science conference. Gain hands-on training in generative AI, LLMs, RAG, AI Safety, and more through expert-led workshops and bootcamps. Explore cutting-edge tools in the AI Expo Hall, connect with industry leaders, and customize your experience with flexible 1- to 4-day passes. Don’t miss this chance to expand your AI skills and network — register now to secure your spot.

Citigroup Pilots AI Agent Capabilities to Automate Complex Workflows

Citigroup is moving closer to a vision of AI agents capable of autonomously handling multi-step corporate workflows, a capability that many enterprises are eager to unlock. The bank announced a pilot program this month that will test “agentic” features within its proprietary AI platform, developed over the past two years, according to The Wall Street Journal.

Autonomous Task Completion with Citi Stylus Workspaces

The new update allows users to issue a single prompt that triggers multiple actions across Citi’s internal systems. According to Chief Technology Officer David Griffiths, this marks a significant shift in reliability for agent-based AI.

“A couple of years ago, you could do agentic things with the early versions of the models that were available then. But they weren’t always very reliable. They weren’t always very good at invoking tools. But they are now,” Griffiths said.

The platform, known as Citi Stylus Workspaces, uses a range of models, including Google’s Gemini and Anthropic’s Claude. One example Griffiths cited: a user can instruct the system to research a client, compile a profile using both public and internal data, and translate the results into another language — all in a single step. Previously, each of these actions required separate user input.

Pilot Scope and Cost Management

The initial pilot will involve 5,000 users over a period of four to six weeks. Griffiths said the goal is to measure usage patterns, impact on productivity, and the cost-to-value ratio. AI agents can become expensive quickly as they consume tokens — units of model usage — especially for longer, more complex tasks.

To address this, Citi has implemented strict cost controls. Griffiths noted that rapidly falling model prices make ROI calculations challenging, but most tasks in this pilot should complete in minutes, keeping costs manageable.

Implications for the Workforce

As the technology matures, Griffiths acknowledged its potential to reshape how work is done.

“Does it mean that we need less people? I don’t know,” he said. “It certainly means that we would get a lot more done. And we’ll see how the workforce evolves with that massive boost of capacity that we’re getting here.”

If the tools advance enough to handle multi-hour or multi-day workflows, they could transform enterprise productivity — and force companies to reevaluate staffing models.

Join thousands of practitioners at ODSC AI West 2025, the leading applied data science conference. Gain hands-on training in generative AI, LLMs, RAG, AI Safety, and more through expert-led workshops and bootcamps. Explore cutting-edge tools in the AI Expo Hall, connect with industry leaders, and customize your experience with flexible 1- to 4-day passes. Don’t miss this chance to expand your AI skills and network — register now to secure your spot.

Aaron Levie on building for the long game in enterprise software at TechCrunch Disrupt 2025

Box CEO and co-founder Aaron Levie joins TechCrunch Disrupt 2025 (Oct 27–29, SF) for a candid fireside chat on what invention really looks like inside a public company, what AI is changing in enterprise software, and the mindset it takes to keep evolving when the tech landscape moves at warp speed.

Regular Bird pricing ends Sept 26 — save up to $668 when you register now

Trump signs executive order to facilitate TikTok deal

Vice President JD Vance said that the deal would value TikTok US at "around $14 billion."

Meta launches ‘Vibes,’ a short-form video feed of AI slop

Think TikTok or Instagram Reels, but every single video you come across is essentially just AI slop.

Viral call-recording app Neon goes dark after exposing users’ phone numbers, call recordings, and transcripts

Call recording app Neon was one of the top-ranked iPhone apps, but was pulled offline after a security bug allowed any logged-in user to access the call recordings and transcripts of any other user.

Microsoft cuts cloud services to Israeli military unit over Palestinian surveillance

The investigation was sparked by a story in The Guardian that reported that Unit 8200, the elite Israel military intelligence unit, was using Azure cloud storage to house data on phone calls obtained through the surveillance of Palestinians in Gaza and the West Bank.

Startup founders say Trump’s $100K H-1B fee is a ‘talent tariff’ that will hurt innovation

The price hike for H1B visas could have detrimental impact on the startup industry, experts say.

Tesla asks EPA not to roll back emissions rules as Trump calls climate change a ‘con job’

The company's regulatory policy team is trying to clean up a mess that CEO Elon Musk spent $300 million to help create.

Spotify denies recent accusation that it changed its terms for artists

Spotify sets the record straight about the distribution rights of artists, podcasters, creators, and authors on the platform.

OpenAI launches ChatGPT Pulse to proactively write you morning briefs

Pulse is part of a broader shift in OpenAI's consumer products, which are lately being designed to work for users asynchronously instead of responding to questions.

Threads is developing a tool that lets you ‘tag’ its algorithm to configure your feed

Threads is developing a new feature that would let users control what they see by 'tagging' the algorithm.

Juicebox raises $30M from Sequoia to revolutionize hiring with LLM-powered search

The recruiting startup has over 2,500 customers, including recruiters at Perplexity, Ramp, and OpenAI.

Steph Curry’s VC firm just backed an AI startup that wants to fix food supply chains

While building AI for food supply chains may sound unglamorous, Burnt argues that decades of failed tech rollouts have left operators skeptical of “tech tourists” with no industry experience.

Amazon to pay $2.5B in FTC settlement over ‘deceptive’ Prime tactics

The company will be required to pay a $1 billion civil penalty and provide $1.5 billion in refunds back to an estimated 35 million consumers harmed by the company's "deceptive Prime enrollment practices," the FTC says.

LGBTQ+ youth have worse mental health outcomes without access to safe online spaces, studies show

These online communities have become more critical in a time LGBTQ+ rights are under attack in the U.S.

OpenAI says GPT-5 stacks up to humans in a wide range of jobs

A new test from OpenAI aims to understand how close AI is to outperforming humans at economically valuable work.

Elon Musk’s xAI offers Grok to federal government for 42 cents

xAI has reached a deal with the U.S. GSA to sell Grok to federal agencies for 42 cents over 18 months, undercutting OpenAI and Anthropic.

Love, lies, and algorithms: Is AI really helping us find ‘the one’? Live at TechCrunch Disrupt 2025

From dating apps and AI-powered matchmaking to full-on digital companionship, artificial intelligence is rapidly becoming a third party in our most personal relationships. But is it truly helping us find deeper connection — or just reshaping romance into an algorithmic illusion?

What top VCs want from AI founders: Inside the investor lens with Jon McNeill, Aileen Lee, and Steve Jang at TechCrunch Disrupt 2025

Jon McNeill (DVx Ventures), Aileen Lee (Cowboy Ventures), and Steve Jang (Kindred Ventures) share what AI founders need to know now: from defensibility to term sheets. TechCrunch Disrupt 2025 takes place October 27–29 in San Francisco. Register before tomorrow ends to save up to $668.

Doorstep raises $8M seed to help find missing food deliveries

Doorstep integrates into existing delivery apps and then, using phone sensors, tracks when a driver has entered a building, gone up an elevator, and made it to the desired doorstep.

Meta rolls out Teen Accounts on Facebook and Messenger globally

Teen Accounts first rolled out to Instagram a year ago after Meta and other popular social networks were grilled by U.S. lawmakers for not doing enough to protect teens on their services.

Why AI isn't replacing radiologists: models underperform in hospital settings, AI use faces legal hurdles, and the job is much more than image recognition (The Works in Progress Newsletter)

The Works in Progress Newsletter:
Why AI isn't replacing radiologists: models underperform in hospital settings, AI use faces legal hurdles, and the job is much more than image recognition  —  Radiology combines digital images, clear benchmarks, and repeatable tasks.  But demand for human radiologists is at an all-time high.

Flox, which lets engineers install a development environment on their machines with a single terminal command, raised a $25M Series B led by Addition (Maria Deutscher/SiliconANGLE)

Maria Deutscher / SiliconANGLE:
Flox, which lets engineers install a development environment on their machines with a single terminal command, raised a $25M Series B led by Addition  —  Flox, a startup that helps software teams create software development environments, has raised $25 million in funding to finance growth initiatives.

Call-recording app Neon, which became the #2 social app on the US App Store, goes dark after exposing users' phone numbers, call recordings, and transcripts (TechCrunch)

TechCrunch:
Call-recording app Neon, which became the #2 social app on the US App Store, goes dark after exposing users' phone numbers, call recordings, and transcripts  —  A viral app called Neon, which offers to record your phone calls and pay you for the audio so it can sell that data to AI companies …

A US federal judge preliminarily approves Anthropic's $1.5B copyright settlement with authors (Blake Brittain/Reuters)

Blake Brittain / Reuters:
A US federal judge preliminarily approves Anthropic's $1.5B copyright settlement with authors  —  A federal judge in California on Thursday preliminarily approved a landmark settlement of a copyright class action brought by a group of authors against artificial intelligence company Anthropic …

Nintendo of America President Doug Bowser will retire on December 31 after more than a decade with the company; Nintendo exec Devon Pritchard will succeed him (Jennifer Maas/Variety)

Jennifer Maas / Variety:
Nintendo of America President Doug Bowser will retire on December 31 after more than a decade with the company; Nintendo exec Devon Pritchard will succeed him  —  Nintendo of America president Doug Bowser will retire Dec. 31 after more than a decade at the company.

Sources: Meta is considering using Google's Gemini and open-source Gemma AI models to improve its ad summarization and recommendation system (Erin Woo/The Information)

Erin Woo / The Information:
Sources: Meta is considering using Google's Gemini and open-source Gemma AI models to improve its ad summarization and recommendation system  —  Meta Platforms staffers have had discussions with Google Cloud about the possibility of using Google's artificial intelligence models to improve Meta's ad business …

Trump says Xi Jinping approved the TikTok deal, and JD Vance says the transaction values TikTok US at $14B; ByteDance hasn't acknowledged the deal (Jonathan Vanian/CNBC)

Jonathan Vanian / CNBC:
Trump says Xi Jinping approved the TikTok deal, and JD Vance says the transaction values TikTok US at $14B; ByteDance hasn't acknowledged the deal  —  President Donald Trump on Thursday signed an executive order approving a proposed deal that would keep TikTok alive in the U.S.

Trump signs an EO declaring that a proposed deal allowing TikTok to continue operating in the US will be a qualified sale that meets national security concerns (Associated Press)

Associated Press:
Trump signs an EO declaring that a proposed deal allowing TikTok to continue operating in the US will be a qualified sale that meets national security concerns  —  President Donald Trump has signed an executive order declaring that a proposed deal allowing TikTok to continue operating …

Google asks SCOTUS to pause a lower court's order requiring a Play Store overhaul set to take effect on October 22, while the company seeks high court review (Peter Blumberg/Bloomberg)

Peter Blumberg / Bloomberg:
Google asks SCOTUS to pause a lower court's order requiring a Play Store overhaul set to take effect on October 22, while the company seeks high court review  —  Alphabet Inc.'s Google asked the US Supreme Court to pause a lower court's order requiring an overhaul of the technology giant's app store policies …

Meta rolls out Vibes, a platform where users can create and share short-form, AI-generated videos, in the Meta AI app and on the meta.ai website (Juby Babu/Reuters)

Juby Babu / Reuters:
Meta rolls out Vibes, a platform where users can create and share short-form, AI-generated videos, in the Meta AI app and on the meta.ai website  —  Meta Platforms (META.O) launched a new feed of AI videos, called Vibes, as the social media giant looks to fast-track work on artificial intelligence technology.

Google updates Gemini 2.5 Flash with better response formatting and image understanding, and releases new 2.5 Flash and 2.5 Flash-Lite previews for developers (Abner Li/9to5Google)

Abner Li / 9to5Google:
Google updates Gemini 2.5 Flash with better response formatting and image understanding, and releases new 2.5 Flash and 2.5 Flash-Lite previews for developers  —  Google today updated the 2.5 Flash model available in the Gemini app across three areas.  —  All responses now benefit …

HSBC says it has tested an IBM quantum computing tool on European bond market data and saw a 34% improvement over traditional methods in predicting order fills (Arjun Neil Alim/Financial Times)

Arjun Neil Alim / Financial Times:
HSBC says it has tested an IBM quantum computing tool on European bond market data and saw a 34% improvement over traditional methods in predicting order fills  —  Europe's largest lender tested a tool developed by IBM on bond market data  —  HSBC claimed quantum computing tools …

Juicebox, whose PeopleGPT helps employers analyze professional profiles to identify qualified candidates, raised $36M, including a $30M Series A led by Sequoia (Marina Temkin/TechCrunch)

Marina Temkin / TechCrunch:
Juicebox, whose PeopleGPT helps employers analyze professional profiles to identify qualified candidates, raised $36M, including a $30M Series A led by Sequoia  —  For years, recruiters used machine learning to find potential hires by searching for keywords in resumes and LinkedIn profiles.

Sources: Intel CEO Lip-Bu Tan has spoken to TSMC CEO C.C. Wei about a partnership or a joint venture; he also met with Tim Cook this year; INTC is up 9%+ (Wall Street Journal)

Wall Street Journal:
Sources: Intel CEO Lip-Bu Tan has spoken to TSMC CEO C.C. Wei about a partnership or a joint venture; he also met with Tim Cook this year; INTC is up 9%+  —  Semiconductor maker has been winning support from other tech companies following the deal that gave the U.S. government 10% stake

Cloudflare announces plans to launch NET Dollar, a US dollar-backed stablecoin designed to support payments for the "agentic web" (Blockworks)

Blockworks:
Cloudflare announces plans to launch NET Dollar, a US dollar-backed stablecoin designed to support payments for the “agentic web”  —  The company introduced a dollar-backed stablecoin to power instant payments and microtransactions for AI-driven web platforms

TUI staff were drowning in HR processes. Here’s how they resurfaced.

TUI found traditional approaches to training staff no longer fit for purpose. A new self-service tool is driving dramatic improvements.

Seven AI takeaways: what we’ve learned about deployment in 2025

A mid-September event in Stockholm, Sweden, marked a break in this year’s Tech Monitor / AMD series of executive leadership events.

Insomniac finally releases the first bloody gameplay trailer for Wolverine

It has been four years since Insomniac announced it was developing a Wolverine game. At the time, Spider-Man 2 was still two years away, which makes it feel like ages ago. What makes it worse is that the studio has been radio-silent since the 2021 PlayStation Showcase reveal – aside...

Read Entire Article

Micron ships 11 Gbps HBM4, confirms 40 Gbps GDDR7 in development

During its Q4 earnings call, Micron claimed that its latest HBM4 stacks deliver pin speeds of more than 11 Gbps and bandwidth exceeding 2.8 TB/s, outperforming competing products in both speed and efficiency. The company attributed these gains to its proven 1-gamma DRAM, a new HBM4 design, its in-house CMOS...

Read Entire Article

Modding project unleashes the full potential of the 3dfx Voodoo graphics architecture

The 3dfx Interactive brand is Gone But Not Forgotten, as modders are still proving with new projects using the chip architecture that made 3D PC gaming possible. YouTuber PixelPipes recently tested a custom Voodoo design by Daniel "sdz" Simionescu, a modder who has been working with Voodoo chips for years....

Read Entire Article

Apple pushes back on EU's Digital Markets Act, aimed at curbing tech giants' power

Apple is pressing European regulators to roll back sweeping rules designed to curb the power of major technology companies, warning that the legislation is weakening consumer protections and stifling innovation rather than enhancing them.

Read Entire Article

Elementary OS offers a polished Linux alternative to Windows and macOS

elementary OS is based on Ubuntu LTS and offers a sleek, beginner-friendly Linux experience. Its signature is designed with simplicity and consistency in mind. Unlike many Linux setups that emphasize extensive customization, Elementary focuses on providing a curated and cohesive user experience that bears a strong resemblance to macOS (in a good way).

The AI bubble is the only thing keeping the US economy together, Deutsche Bank warns

According to a research note recently sent to clients by Deutsche Bank, the AI boom is currently helping the US economy avoid a recession but it cannot continue indefinitely. George Saravelos, Global Head of FX Research at Deutsche Bank, said the US would be close to a recession this year...

Read Entire Article

Sony's Pulse Elevate wireless speakers are built for PC gamers, powered by planar magnetic drivers

The speakers, announced during Sony's recent State of Play presentation, feature planar magnetic drivers, built-in woofers, and a microphone in the right speaker for voice chat and AI-enhanced noise rejection. The integrated rechargeable battery affords "hours" of battery life, and charging docks are included as part of the bundle.

Read Entire Article

Samsung Galaxy Z Flip 7 FE, the budget-friendly foldable, hits record low at $699

The Galaxy Z Flip 7 FE features a 6.7-inch 120 Hz foldable AMOLED display and is powered by the Exynos 2400 SoC with 8 GB of RAM and 128 GB of storage. Recently launched and reviewed, critics praise it for matching its pricier counterpart in camera and performance quality. Normally $899, it's now available at a record-low $699.

Windows ML debuts for all developers, enabling AI apps to run directly on local hardware

After its original introduction at Build 2025, Microsoft has now made Windows ML generally available to all developers targeting Windows 11 24H2. The technology provides a runtime with built-in AI inferencing capabilities, optimized for on-device model execution. Microsoft hopes the solution will help accelerate the flood of AI products now...

Read Entire Article

Intel turns to Apple for potential investment as it looks to rebuild chip business

Intel has reportedly held early talks with Apple about a possible investment. The struggling chipmaker is seeking to stabilize its business with new backers following recent partnerships with other rivals and overseas investors. Anonymous sources told Bloomberg that the discussions remain preliminary and may not result in a deal.

Read Entire Article

Qualcomm pushes hybrid AI forward with Snapdragon X2 Elite and 8 Elite Gen 5

If you're paying close attention, you can't help but notice a key theme that's started to weave its way into and through the messaging from today's top tech companies. That theme? Hybrid AI - a combination of running AI models, agents, and applications both in the cloud and on the...

Read Entire Article

Deus Ex Remastered comes to PC and consoles February 5, bringing visual and gameplay upgrades

Aspyr Media unveiled an upcoming remaster of Ion Storm's seminal classic, Deus Ex, during Sony's State of Play showcase at the 2025 Tokyo Game Show. The game will be available for $29.99 on Steam, PlayStation 5, Xbox, and Nintendo Switch next February.

Read Entire Article

China's latest Fenghua GPU debuts with 112+GB of HBM memory, claims CUDA compatibility and ray tracing support

The Fenghua No. 3 is based on the open-source RISC-V architecture, with inputs from the OpenCore Institute's Nanhu V3 project. The new design is expected to make the card more powerful and efficient than its predecessors - Fenghua No. 1 and No. 2 - both of which were based on...

Read Entire Article

Infinite Machine's Cybertruck-like P1 e-scooter is launching soon

Infinite Machine's P1 looks like something you'd see in a movie from the 1980s set in the future. Announced earlier this year, the vehicle is made from brushed aluminum and steel monocoque. It features a 7-inch touchscreen, a speaker, and comes with an app for unlocking and starting the P1....

Read Entire Article

TSMC reports progress on A14 node with 15% performance boost and 30% power reduction

TSMC has released new technical details on its A14 semiconductor process, reporting on both the performance and efficiency gains expected to surpass its 2-nanometer platform. The company said development of the node is ahead of schedule, signaling continued progress even as much of the industry struggles with ramping up current-generation...

Read Entire Article

Steam beta helps players prep for games requiring Secure Boot or TPM

The only new feature included in the September 23 Steam client beta update allows users to see whether their PCs support Secure Boot and TPM from within the launcher. The minor addition will help users who are preparing to purchase or launch certain online games.

Read Entire Article

Forza Horizon 6 set in Japan, coming to PC and Xbox first in 2026

There has been a slew of rumors claiming that the next Forza Horizon game would be set in Japan. Aptly, it was at the Xbox showcase at the Tokyo Game Show where Microsoft confirmed Forza Horizon 6 would indeed be set in the land of the rising sun.

Read Entire Article

San Francisco shuts down website that helped drivers avoid parking tickets – four hours after launch

Find My Parking Cops, which imitates Apple's Find My Friends app in both name and design, let users see the locations of parking enforcement officials on a map in near real-time.

Read Entire Article

Android users can now edit photos using voice commands with Gemini AI

Google is rolling out a new AI photo editing tool for Android users, allowing them to modify images in Google Photos using conversational commands. The feature began launching this week and introduces a natural language interface that simplifies photo editing, eliminating the need to navigate a series of editing menus...

Read Entire Article

The Complete List of Alternatives to Every Google Product

Discover privacy friendly alternatives to every Google product. Take small steps to protect your data, reduce tracking, and regain control of your digital life with secure, reliable tools.

Is LLM in-context learning real learning? A Microsoft study says yes, but it’s not what you think

In-context learning is learning, but it's a brittle, superficial process that relies more on patterns than true understanding.

The post Is LLM in-context learning real learning? A Microsoft study says yes, but it’s not what you think first appeared on TechTalks.

Even Disney reportedly lacks enough data to train a top-tier AI video model

Lionsgate's AI deal with startup Runway is moving slower than expected.

The article Even Disney reportedly lacks enough data to train a top-tier AI video model appeared first on THE DECODER.

Open source Qwen3-VL outperforms Gemini 2.5 Pro in major vision benchmarks, Alibaba reports

Alibaba has released Qwen3-VL, an open source language vision model that works with both images and text.

The article Open source Qwen3-VL outperforms Gemini 2.5 Pro in major vision benchmarks, Alibaba reports appeared first on THE DECODER.

Microsoft reportedly plans AI content marketplace for select publishers

Microsoft is developing a pilot project called the "Publisher Content Marketplace" (PCM).

The article Microsoft reportedly plans AI content marketplace for select publishers appeared first on THE DECODER.

OpenAI's new Pulse feature lets ChatGPT start the conversation

OpenAI is testing a new feature called Pulse that aims to make ChatGPT a more proactive assistant. Instead of waiting for questions, ChatGPT now pulls in information on its own and delivers daily updates.

The article OpenAI's new Pulse feature lets ChatGPT start the conversation appeared first on THE DECODER.

Meta's Code "World Model" aims to close the gap between code generation and code understanding

Meta's Code World Model (CWM) is designed not just to generate code but to understand how that code runs on a computer.

The article Meta's Code "World Model" aims to close the gap between code generation and code understanding appeared first on THE DECODER.

Alibabas Wan2.5-Preview lets users turn photos and text prompts into videos with matching audio

Alibaba has launched Wan2.5-Preview, a new video model capable of generating short clips with synchronized audio.

The article Alibabas Wan2.5-Preview lets users turn photos and text prompts into videos with matching audio appeared first on THE DECODER.

OpenAI seeks advertising lead to oversee monetization for ChatGPT

OpenAI is hiring a manager to lead advertising efforts for ChatGPT.

The article OpenAI seeks advertising lead to oversee monetization for ChatGPT appeared first on THE DECODER.

OpenAI, Oracle and Softbank announce five more Stargate Centers

OpenAI, Oracle, and SoftBank are scaling up their Stargate AI infrastructure platform across the US.

The article OpenAI, Oracle and Softbank announce five more Stargate Centers appeared first on THE DECODER.

Okta Wants To Secure Your AI Agents, Too

At its annual Oktane user conference, identity management service Okta today announced its plans to start treating AI agents —

The post Okta Wants To Secure Your AI Agents, Too appeared first on The New Stack.

JavaScript Gets Supply Chain Security With Chainguard Libraries

is expanding its supply chain security platform with the launch of Chainguard Libraries for JavaScript, a collection of thousands of

The post JavaScript Gets Supply Chain Security With Chainguard Libraries appeared first on The New Stack.

How To Use RAG-Powered AI for Safer Legacy Code Maintenance

The reality is that generative AI (GenAI) was never “introduced” into enterprise business; it crept in when users discovered it

The post How To Use RAG-Powered AI for Safer Legacy Code Maintenance appeared first on The New Stack.

Microsoft Launches a Combined Marketplace for Cloud Solutions, AI Apps and Agents

Microsoft has long offered a variety of marketplaces for businesses in its ecosystem. For cloud solutions, there is the Azure

The post Microsoft Launches a Combined Marketplace for Cloud Solutions, AI Apps and Agents appeared first on The New Stack.

Consistency at Scale: Unifying Temporal and YugabyteDB

AI, reliability and governance aren’t optional — they’re the foundation on which everything else depends. Data platform Manetu unified Temporal

The post Consistency at Scale: Unifying Temporal and YugabyteDB appeared first on The New Stack.

Why Your App’s Biggest Performance Bottleneck Might Be SSL/TLS

Your app is slow. You’ve optimized the database, fixed the N+1 queries and tuned your microservices. But what if the

The post Why Your App’s Biggest Performance Bottleneck Might Be SSL/TLS appeared first on The New Stack.

The Cross App Access Protocol Makes AI Agents Enterprise-Ready

LAS VEGAS — Most enterprises are now using AI agents in some form, but few have any governance systems in

The post The Cross App Access Protocol Makes AI Agents Enterprise-Ready appeared first on The New Stack.

AI that once called itself MechaHitler will now be available to the US government for $0.42

Elon Musk's AI appears to be more ideological than competitors

Despite protest letters, concerns that it's biased and untrustworthy, model tweaks to appease its billionaire boss, and even a past incident where it called itself "MechaHitler," xAI's Grok is still being made available to government agencies for mere pennies.…

Harness pitches AI agents as your new DevOps taskmasters

Productivity gains promised, but humans still expected to audit the bots

At its Unscripted event in London, DevOps company Harness presented its latest AI-driven modules, including an AI pipeline builder, AI test automation, autonomous code fixing when builds fail, AI AppSec (application security) and even AI-driven chaos testing, where resiliency is tested by introducing random failures.…

DARPA wants AI to know when it's being an energy hog

New research program seeks ‘energy-aware’ ML that balances performance with power draw

It's notoriously difficult to consistently measure the energy usage of AI models, but DARPA wants to put an end to that uncertainty with new "energy-aware" machine learning systems. …

Oracle saddles up with $18B debt amid AI infrastructure gamble

Ballooning leverage and shaky customer funding could strain Big Red's balance sheet

Oracle has raised $18 billion in debt, which could help fund massive datacenter investments aimed at meeting surging demand from AI model builders and enterprise customers.…

90% of devs now use AI (but don't trust it)

PLUS: Make pixel-perfect website changes with AI

The Sequence Opinion #726: The Shock Alliance: Nvidia × Intel Rewires the Rack

One of the most surprising events in the world of generative AI.

GCEX Group acquires GlobalBlock to drive digital asset expansion

The deal is set to support the digital prime broker in expanding from its current OTC, conversion and technology business towards a wider digital asset and wealth management business. 

The post GCEX Group acquires GlobalBlock to drive digital asset expansion appeared first on The TRADE.

Nasdaq expands partnership to allow institutions to deploy Calypso on AWS

The offering will allow clients to continuously operate on the most advanced service and forms part of the firms’ drive to modernise capital markets and treasury infrastructure.  

The post Nasdaq expands partnership to allow institutions to deploy Calypso on AWS appeared first on The TRADE.

Mark Freeman makes industry return with xyt director role

Individual returns three years after having announced his retirement back in 2022 following 42 years in the industry.  

The post Mark Freeman makes industry return with xyt director role appeared first on The TRADE.

Former Clear Street managing director emerges at JP Morgan in equity sales trading role

Individual joins the firm following his departure from Clear Street in July 2025, after the broker announced several cuts to its UK team months after appointments.  

The post Former Clear Street managing director emerges at JP Morgan in equity sales trading role appeared first on The TRADE.

Nintendo of America boss Doug Bowser is retiring

One of Nintendo’s most recognizable names is stepping down. Today the company announced that the aptly named Doug Bowser, who took over as president of Nintendo of America from Reggie Fils-Aimé in 2019, will be retiring. He’ll stay on until the end of the year — December 31st, to be precise — and will be […]

Nvidia is letting anyone use its AI voice animation tech

Nvidia is open-sourcing Audio2Face, its AI-powered tool that generates realistic facial animations for 3D avatars — all based on audio input. The change means developers can now use the tool and its underlying framework to create realistic 3D characters for their games and apps. Nvidia’s Audio2Face works by analyzing the “acoustic features” of a voice, […]

Trump signs ‘Saving TikTok’ order to start resolving its big ban problem

President Donald Trump has signed an executive order recognizing the framework of a deal between ByteDance and the US that would satisfy the TikTok divest-or-ban law. The deal values TikTok’s US operations at $14 billion and puts it under the control of companies based in the US. “I spoke with President Xi [Jinping], we had […]

Silicon Valley’s latest argument against regulating AI: that would literally be the Antichrist

In a four-part series of religious lectures in San Francisco, Peter Thiel - yes, that Peter Thiel - has argued that the End Times are nigh and that a biblical Antichrist - yes, that Antichrist - will come to Earth in the form of onerous government regulations placed on science, technology, and AI. These are, […]

Microsoft says this new cooling method could enable more powerful chips and efficient data centers

Microsoft is making advances with a new way to cool microchips that it says could lead to more energy-efficient data centers in the future. It's a method called microfluidics that involves liquid coolant flowing directly into the silicon. After lab tests, Microsoft found that this strategy can remove heat up to three times better than […]

A number of great chargers are already on sale ahead of October Prime Day

Amazon’s October Prime Day event (also known as Prime Big Deal Days) doesn’t officially start until October 7th, but that doesn’t mean you have to wait to find a good deal on charging accessories. Many speedy wall adapters, power banks, and large battery backup solutions are already on sale now from brands like Anker and […]

This Energizer laptop claims to have a battery that keeps going and going and going

Avenir Telecom is making a line of new Energizer-branded laptops aimed at marathon battery life. There are three productivity-focused models: the Energizer EnergyBook Pro 15, EnergyBook Pro XL 18, and flagship Energizer EnergyBook Pro Ultra. The Pro Ultra sports a giant 18-inch display with a meager 1920 x 1200 resolution and an absolutely massive 192Wh […]

Twitch is making it easy to rewind streams

Twitch is launching a feature this week I’ve wanted for a very long time: the ability to scrub back on a livestream right from a streamer’s main channel page. You can already look back at previous parts of livestreams as they’re going on by clicking through to the Videos section of a streamer’s profile, but […]

Gemini now explains why your Sheets formula failed

Nine months after Google infused Gemini AI into Google Sheets, the AI system graduated from text and charts to taking on formulas. (Even those plopped into messy wedding planning spreadsheets, I found.)  Gemini’s chatbot appears on the right-hand side of Sheets. Now, when you ask how to manipulate data, Gemini will respond with suggested formulas […]

What happens when an AI-generated artist gets a record deal? A copyright mess

Two weeks ago, record company Hallwood Media signed a deal with Telisha "Nikki" Jones after negotiations that purportedly included an offer of $3 million, Billboard reported. Jones is a Mississippi-based lyricist behind the R&B artist "Xania Monet" whose most popular song on Spotify racked up over 1 million listens, and whose Reels regularly top 100,000 […]

Prototype-Based Models and The Growing Importance of Interpretable AI

Artificial Intelligence (AI) has transformed how we approach problems in science, industry, and everyday life. Deep learning models now power everything from medical image analysis to autonomous vehicles. While these models deliver remarkable accuracy, they often come with a major drawback: they are black boxes. Their inner workings are so complex that even experts struggle to explain why a particular decision was made.

This lack of transparency is not just an academic issue — in high-stakes fields such as healthcare, law, finance, and biotechnology, understanding the reasoning behind a model’s decision is just as important as the decision itself. Trust, accountability, and fairness all depend on interpretability.

Traditional interpretability methods fall short

To address the opacity of deep neural networks, a variety of post-hoc interpretability techniques have been developed. Methods such as saliency maps[1] and Grad-CAM[2] visualize which parts of an input image most influenced a model’s decision. These techniques are intuitive and have been widely adopted, but they have significant limitations:

• They often highlight “where” a model is looking, but not why that region matters for classification.
• The explanations are not tied to the actual decision-making process of the model, making them unreliable in practice.
• Small perturbations to the input can dramatically change the explanation, reducing trust in their stability.

As a result, researchers have sought approaches that go beyond these approximations and build interpretability into the architecture of the model itself.

Enter prototype-based models

Prototype-based models represent a promising direction in interpretable machine learning. Instead of treating explanations as an afterthought, these models make predictions in a case-based reasoning framework:

• The model learns prototypes — small patches or exemplars from the training data that capture distinctive features of each class.
• Predictions are made by comparing parts of a new input to these prototypes.
• Explanations come naturally: the model can say, “This image looks like that prototypical example,” mimicking the way humans justify decisions.

This approach offers faithful explanations because the prototypes are directly tied to the model’s internal reasoning, unlike saliency maps or heatmaps that are generated after the fact.

Key prototype-based architectures

The field of prototype networks has evolved quickly, with several notable architectures addressing different aspects of interpretability

ProtoPNet: The First Step Toward Prototype-Based Interpretability

The first major breakthrough in prototype-based interpretability was ProtoPNet [4]. This architecture was the foundation for much of the subsequent work in the field.

ProtoPNet builds on a standard convolutional neural network by adding a prototype layer. Instead of making predictions directly from feature maps, the model learns a fixed number of part-prototypes for each class. These prototypes correspond to meaningful image patches (e.g., a bird’s wing or a car’s headlights) and are compared to regions of a new input image using $L_2$ distance. A prediction is then made based on how strongly the input matches the learned prototypes.

To generate explanations, ProtoPNet projects each prototype back onto the closest patch from the training set. This allows the model to justify its decision in human terms: “This part of the image looks like that part of a training example.”

Why it mattered

This case-based reasoning marked a major shift in interpretable deep learning. Instead of relying on heatmaps or post-hoc approximations, ProtoPNet tied explanations directly to the way the model makes predictions. The result was a framework that was not only accurate but also transparent and intuitive. The key note here is that ProtopNet uses latent space representations of actual training images as prototypes in a step called ‘prototype projection’. This allows the user to visualize the prototypical parts in image space and verify whether the prototype is truly representative of the visual feature.

Limitations

Despite its novelty, ProtoPNet also exposed important challenges that shaped later research:

• Prototype inconsistency: The same prototype could activate on different object parts across images, reducing explanation reliability.
• Prototype instability: Small input perturbations (like noise) could cause prototypes to shift activations, undermining robustness.
• Poor diversity: Multiple prototypes often collapsed onto the same visual feature, limiting the richness of explanations.

These issues highlighted the need for more reliable, stable, and diverse prototypes — problems that inspired many follow-up models such as TesNet[7], Deformable ProtoPNet[6] and ProtoPAligned[5].

Other notable architectures

Building on ProtoPNet, several extensions have been proposed to address limitations in prototype diversity and reliability.

TesNet introduced cosine similarity in place of Euclidean distance and incorporated an orthogonality loss, encouraging prototypes to be both more distinct and more consistent.

Following this direction, models such as Deformable ProtoPNet and ProtoPool further adopted cosine similarity and orthogonality loss to enhance prototype diversity and improve overall performance.

ProtoPAligned shifted the focus more explicitly toward interpretability by adding architectural modules such as Shallow–Deep Feature Alignment and Score Aggregation. It also formally introduced the consistency and stability scores, establishing quantitative metrics for evaluating prototype reliability and marking an important move away from purely qualitative inspection.

Why prototype-based models matter

Prototype networks stand out because they provide case-based explanations directly tied to the decision-making process. This is particularly powerful in fine-grained tasks such as distinguishing bird species, medical diagnoses, or defect detection, where small, localized differences matter.

By grounding predictions in real examples, these models offer:

• Faithfulness: Explanations reflect how the model actually makes decisions.
• Transparency: Users can see which parts of an input are matched with meaningful prototypes.
• Trustworthiness: Stable and consistent prototypes help build confidence in the model’s reasoning.

At the same time, challenges remain. Ensuring prototype diversity, robustness to noise, and scalability to large datasets are ongoing research questions. Nonetheless, the trajectory of work in this field demonstrates an encouraging trend: interpretability is being treated as a first-class goal, not an afterthought.

Closing thoughts

Prototype-based models are reshaping how we think about interpretable machine learning. From the early breakthroughs of ProtoPNet to the more advanced formulations of ProtoPAligned and beyond, these models provide a blueprint for designing systems that are not only accurate but also interpretable.

As AI continues to move into critical domains, the importance of such approaches cannot be overstated. While no single model has solved interpretability, prototype-based networks are a step toward bridging the gap between black-box performance and human-centered transparency — a step that may ultimately make AI more accountable, trustworthy, and useful in the real world.

References:

1. K. Simonyan, “Deep inside convolutional networks: Visualising image classification models and saliency maps,” arXiv preprint arXiv:1312.6034, 2013

2. R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-cam: Visual explanations from deep networks via gradient-based localization,” in Proceedings of the IEEE international conference on computer vision, pp. 618–626, 2017.

3. T. Laugel, M.-J. Lesot, C. Marsala, X. Renard, and M. Detyniecki, “The dangers of post-hoc interpretability: Unjustified counterfactual explanations,” arXiv preprint arXiv:1907.09294, 2019.

4. C. Chen, O. Li, C. Tao, A. J. Barnett, J. Su, and C. Rudin, “This looks like that: Deep learning for interpretable image recognition,” 2018.

5. Q. Huang, M. Xue, W. Huang, H. Zhang, J. Song, Y. Jing, and M. Song, “Evaluation and improvement of interpretability for self-explainable part-prototype networks,” tech. rep., 2023.

6. J. Wang, H. Liu, X. Wang, and L. Jing, “Interpretable image recognition by constructing transparent embedding space,” in Proceedings of the IEEE/CVF international conference on computer vision, pp. 895–904, 2021.

7. J. Donnelly, A. J. Barnett, and C. Chen, “Deformable protopnet: An interpretable image classifier using deformable prototypes,” 2024.

Prototype-Based Models and The Growing Importance of Interpretable AI was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

4 Context Engineering Techniques to Create Powerful LLM Applications

Stochastic Pathways of Long-Term Investing: A Control, Learning, and Search Perspective

LAI #94: Deep Learning Myths, Multi-Agent Frameworks, and Synthetic Data in Practice

From AutoGen workflows to linear estimators, self-driving cars, and principles for multi-agent design.

Good morning, AI enthusiasts,

This week, we take a closer look at what deep learning really is — and what it isn’t. Rather than true intelligence, it’s better thought of as sophisticated pattern-matching, which helps explain both its remarkable successes and its blind spots when the data shifts.

We then turn to the systems built on top of these models. One deep dive explores how to structure multi-agent teams with AutoGen, from managing group conversations to optimizing workflows. Another lays out the key principles for designing agent systems that scale without breaking down. Alongside these, you’ll find a clear guide to linear estimators as a tool for balancing uncertainty, and a breakdown of the five layers of complexity behind self-driving cars, from perception to control.

Together, these pieces highlight the current limitations of AI while demonstrating how thoughtful system design can expand what’s possible.

Let’s get into it.

What’s AI Weekly

https://medium.com/media/7bce5ef231a9a7e79c5cd15394aa1eb6/href

This week, in What’s AI, I break down what deep learning really is and why it’s not the same as intelligence. Think of it less as “understanding” and more as copying patterns, like a student guessing the most common answers. We’ll see where this works brilliantly, like generating text or analyzing images, and where it fails once the data shifts. I also share why this gap matters if we want to use AI responsibly. Read the full article here or watch the video on YouTube.

— Louis-François Bouchard, Towards AI Co-founder & Head of Community

Learn AI Together Community Section!

Featured Community post from the Discord

Cristhiangomez_18174 is building Handit, a teammate for your AI agents that makes them production-ready with just one command. Users can monitor every interaction, detect issues in real-time, and auto-generate PRs with tested fixes. He has shared a short tutorial showing how to connect it to a LangGraph agent (from LangChain). Read the article here and get the tool. If you have any questions or feedback, connect with him in the thread!

AI poll of the week!

Most of you see synthetic data not as the future of training, but as a niche tool for specific contexts. That matches the industry trend: big labs like OpenAI and Anthropic experiment with synthetic corpora to stretch models further, but when it comes to production, synthetic data is usually reserved for patching gaps — rare edge cases, safety-critical scenarios, or privacy-sensitive domains.

As datasets get scarcer and regulations tighten, will the role of synthetic data expand, or will it always remain a supplement to real-world data? Share in the thread!

Collaboration Opportunities

The Learn AI Together Discord community is flooding with collaboration opportunities. If you are excited to dive into applied AI, want a study partner, or even want to find a partner for your passion project, join the collaboration channel! Keep an eye on this section, too — we share cool opportunities every week!

1. Andreas532707 is looking for a partner to build an AI project. If you are a beginner starting your first project, connect with him in the thread!

2. Armodr is looking to connect with someone in Southeast Asia to discuss AI automation/AI agency with low-to-no code AI tools. Connect with him in the thread to know more.

3. Bavoyager is looking for a developer with experience in AI, video & audio transcription for a project. If this sounds like you, find more details in the thread.

Meme of the week!

Meme shared by efficientnet_99825

TAI Curated Section

Article of the week

Building Multi-Agent Teams with AutoGen: Deep Dive By Aayushi_Sharma

Complex tasks often demand more than a single AI agent. This article shows how to build and manage multi-agent systems with Microsoft’s AutoGen framework. It focuses on two structures: RoundRobinGroupChat, which runs sequential, turn-based conversations, and SelectorGroupChat, which delegates tasks to the best-fit agent. Through practical code examples, it demonstrates how to oversee the team lifecycle, from real-time monitoring to setting clear termination conditions, so workflows stay efficient and under control.

Our must-read articles

1. A Simple (But Not Too Simple) Intro to Linear Estimators By Maxwell’s Demon

This article addresses the common challenge of combining prior knowledge with a new, potentially noisy measurement. It introduces linear estimators as a clean mathematical solution. The method produces a weighted average, balancing the uncertainty of old and new information. With clear examples (such as temperature readings), the author demonstrates how strong priors resist noisy updates — and connects the concept to more advanced tools like Kalman filters.

2. The Complexity of Self-Driving Cars Explained Simply By Maxwell’s Demon

Self-driving cars must replicate the split-second judgments of human drivers. This piece breaks the challenge into five steps: journey information, perception, prediction, planning, and control. It explains the tech behind each — from LiDAR fusion to ML-based prediction and control theory. It also clarifies the SAE’s six levels of autonomy, grounding today’s progress against the goal of full Level 5 automation. Along the way, it surfaces the technical, infrastructural, and ethical hurdles that still stand in the way.

3. Multi-Agent Systems Done Right By Vlad Johnson

This article details principles for designing effective multi-agent AI systems. It outlines key principles: place capable LLMs in supervisory roles with complete context, define explicit success and failure criteria to avoid loops, and build simple hierarchical structures before scaling. It also explores combining different model families to leverage complementary strengths and utilizing long-term memory for more diverse outputs. The piece concludes with a comparison of popular frameworks, providing practical guidance for teams building at scale.

If you are interested in publishing with Towards AI, check our guidelines and sign up. We will publish your work to our network if it meets our editorial policies and standards.

LAI #94: Deep Learning Myths, Multi-Agent Frameworks, and Synthetic Data in Practice was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Explaining Tongyi DeepResearch

Toward the Era of Synthesized Data Training

What will the future of LLM training be like?

Last week, Alibaba Tonyi Lab released its agentic research model, Tongyi DeepResearch, which outperformed OpenAI o3 and DeepResearch on various tasks. Moreover, Tongyi DeepResearch only has 30B parameters in total, with 3B activated per token, while its open-source rival - the DeepSeek v3.1 has 671B parameters, and Kimi Researcher (based on Kimi v2) has 1T parameters. We can’t help but ask, how did Tongyi DeepResearch achieve that?

It turns out not to be magic. Extending the basic ReAct reasoning paradigm to the Iterative Deep Research Paradigm, the model fully utilized synthetic trajectory data for training purposes. Besides the official blog, the synthetic and training technologies are outlined in the following research papers:

• AgentFounder for model pretraining — synthesizing first and higher-order action data for continual pretraining
• WebSailor-V2 and WebShaper for post-training — using random walk for QA knowledge graph dataset creation, and layer-wise expansion-based information seeking strategy for improving data quality, for SFT and GRPO-based RL.

We’ll explain these strategies in detail in the following.

AgentFounder — Agentic Continual Pretraining

The AgentFounder training schema proposed in the paper “Scaling Agents via Continual Pre-training” involves two stages: i) Stage I pretraining with 32K context length, and ii) Stage II with 128K context length.

To create the corresponding agentic data, two different synthesis methods were involved: a) first-order action synthesis (FAS) with planning action and reasoning action; b) higher-order action synthesis (HAS) for multi-decision action synthesis.

First-order action synthesis. According to the original paper, different from traditional “wiki-style” knowledge representation, such as “Paris is the capital of France”, the knowledge is anchored by “entity” France, e.g., (“France”: “Tourist arrivals in France reached 4,222 thousand in June 2025”), to form a diverse QA set.

Higher-order action synthesis. Instead of using the original “plain” reasoning/action trajectory data, a new reasoning/action candidate set is scaled (created) for each step by LLMs so as to help explore different decision possibilities, without changing the final binary decision.

WebSailor-V2 — Creating Knowledge Graph for SFT and RL

The original WebSailor paper proposed a three-level web-scraping method, which progressively built the knowledge graph:

The WebSailor v2 further creates cyclic graphs to introduce information interconnectivity and then uses random walks for subgraph extraction.

Post Training — SFT Cold Start + RL with GRPO. The QA data generated using the pruned knowledge graph will eventually be utilized in the post-training stages, in both the SFT and RL stages.

The reinforcement learning algorithm is GRPO, which DeepSeekMath proposed for improving the PPO by replacing the value model with group scores.

WebShaper — An Information-driven Paradigm for Improving QA Quality

Since the WebSailor v2 first retrieves the knowledge graph from the web and then generates the QA dataset (Information-driven), there could be an inconsistency between the knowledge graph and the final reasoning dataset. To prevent this from happening, the WebShaper paper proposed a “knowledge projection” operation and used set theory to control the structure of the newly generated QA set (Formation-driven).

First, to achieve knowledge/reasoning consistency, WebShaper proposed projecting entities based on certain relationships onto a subdomain using union and intersection operations.

Second, starting from a seed question, a “Layer-wise” structure is created by a sequence of Agentic expansions (composed of union and intersection mentioned above), which eliminates the redundancy and reasoning shortcut issues during QA creation.

ReAct and IterResearch Inference Paradigm

So, how do we integrate the above-created synthetic dataset for agentic inference purposes? Tongyi DeepResearch supports two paradigms: ReAct and IterResearch (or heavy mode).

ReAct is the paradigm proposed in ICLR’23, which follows the “thought-action-observation” cycle as follows:

IterResearch employs a more complex paradigm, which creates a new workspace with a central report that includes noise prevention and quality checks. If a tool call is required, it returns the corresponding tool response from preset environments, along with the predicted action for iteration.

Final Results and Takeaways

How well does the Tongyi DeepResearch perform? The following is their official benchmark results:

The Tongyi DeepResearch model outperforms OpenAI DeepResearch by 6.3% on the Humanity’s Last Exam dataset, a multimodal benchmarking dataset comprising frontier knowledge from mathematics, physics, social science, and other fields. However, the Tongyi DeepResearch lags on the BrowseComp dataset, which “require persistently navigating the internet in search of hard-to-find, entangled information”. This might be because the Tongyi model has a small base of only 30B and only supports a context length of 128k.

But this shows promise. With a small to medium-sized base, the Tongyi DeepResearch model shows great potential for scaling with synthetic trajectory data. It may reveal that we’re facing a completely different era of new training paradigms for synthetic data.

References

• Blog post, “Tongyi DeepResearch: A New Era of Open-Source AI Researchers”, https://tongyi-agent.github.io/blog/introducing-tongyi-deep-research/
• Blog post, “Kimi K2: Open Agentic Intelligence”, https://moonshotai.github.io/Kimi-K2/
• Model card, “DeepSeek-V3.1-Terminus”, https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus
• Tao Z, Wu J, Yin W, Zhang J, Li B, Shen H, Li K, Zhang L, Wang X, Jiang Y, Xie P. Webshaper: Agentically data synthesizing via information-seeking formalization. arXiv preprint arXiv:2507.15061. 2025 Jul 20.
• Li K, Zhang Z, Yin H, Ye R, Zhao Y, Zhang L, Ou L, Zhang D, Wu X, Wu J, Wang X. WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning. arXiv preprint arXiv:2509.13305. 2025 Sep 16.
• Su L, Zhang Z, Li G, Chen Z, Wang C, Song M, Wang X, Li K, Wu J, Chen X, Qiao Z. Scaling Agents via Continual Pre-training. arXiv preprint arXiv:2509.13310. 2025 Sep 16.
• Li K, Zhang Z, Yin H, Zhang L, Ou L, Wu J, Yin W, Li B, Tao Z, Wang X, Shen W. WebSailor: Navigating Super-human Reasoning for Web Agent. arXiv preprint arXiv:2507.02592. 2025 Jul 3.
• Shao Z, Wang P, Zhu Q, Xu R, Song J, Bi X, Zhang H, Zhang M, Li YK, Wu Y, Guo D. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300. 2024 Feb 5.
• Phan L, Gatti A, Han Z, Li N, Hu J, Zhang H, Zhang CB, Shaaban M, Ling J, Shi S, Choi M. Humanity’s last exam. arXiv preprint arXiv:2501.14249. 2025 Jan 24.
• Wei J, Sun Z, Papay S, McKinney S, Han J, Fulford I, Chung HW, Passos AT, Fedus W, Glaese A. Browsecomp: A simple yet challenging benchmark for browsing agents. arXiv preprint arXiv:2504.12516. 2025 Apr 16.

Explaining Tongyi DeepResearch was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Querying APIs with Graph Intelligence: Agents That Truly Understand

TDS Newsletter: To Better Understand AI, Look Under the Hood

AI-powered tools tend to generate extreme reactions: on one side we have the “It’s magic!” and “best thing ever!” crowd. On the other, we find the “we’re doomed!” camp. These aren’t static or monolithic groups, of course. You might even find yourself on both ends of the spectrum — in the span of a single day.  […]

The post TDS Newsletter: To Better Understand AI, Look Under the Hood appeared first on Towards Data Science.

Notes on LLM Evaluation

A practical, step-by-step guide to building an evaluation pipeline for a real-world AI application

The post Notes on LLM Evaluation appeared first on Towards Data Science.

Building a Video Game Recommender System with FastAPI, PostgreSQL, and Render: Part 2

Deploying a FastAPI + PostgreSQL recommender system as a web application on Render

The post Building a Video Game Recommender System with FastAPI, PostgreSQL, and Render: Part 2 appeared first on Towards Data Science.

Building Video Game Recommender Systems with FastAPI, PostgreSQL, and Render: Part 1

Designing a video game recommendations service with Steams API

The post Building Video Game Recommender Systems with FastAPI, PostgreSQL, and Render: Part 1 appeared first on Towards Data Science.

Copilot Does Azure-Focused .NET App Modernizations in Visual Studio

GitHub Copilot app modernization is now generally available in Visual Studio, providing AI-powered upgrades and Azure migration for .NET apps, with automated remediation, CVE scanning, and containerization support.

AI-powered Auto-Categorization now available in Microsoft Photos

Hello Windows Insiders,

We’re excited to announce an update to the Microsoft Photos app, now rolling out on Windows 11 across all Insider channels. This release brings a powerful new feature on Copilot+ PCs that leverages AI to organize photos int

The post AI-powered Auto-Categorization now available in Microsoft Photos appeared first on Windows Blog.

Announcing Windows 11 Insider Preview Build 27954 (Canary Channel)

Hello Windows Insiders, today we are releasing Windows 11 Insider Preview Build 27954 to the Canary Channel.

The post Announcing Windows 11 Insider Preview Build 27954 (Canary Channel) appeared first on Windows Blog.

Unlock text editing use cases with highlightsFromPoint and other FromPoint APIs

We're excited to announce the availability of the highlightsFromPoint() method in Microsoft Edge 140, and other Chromium-based browsers, a useful addition to the CSS Custom Highlight API, which makes it possible to f

The post Unlock text editing use cases with highlightsFromPoint and other FromPoint APIs appeared first on Windows Blog.

Japanese developers bring legendary craftsmanship to newest gaming handhelds

When Xbox debuted in 2001, it was a newcomer to a scene dominated by Japanese-made games and consoles. At that time, Sega, Nintendo and Sony were household names to anyone who’d ever touched a joystick or controller.

Xbox would eventually catch up

The post Japanese developers bring legendary craftsmanship to newest gaming handhelds appeared first on Windows Blog.

RoboCup Logistics League: an interview with Alexander Ferrein, Till Hofmann and Wataru Uemura

The Industrial League arena at RoboCup2025. RoboCup is an international scientific initiative with the goal of advancing the state of the art of intelligent robots, AI and automation. The annual RoboCup event took place from 15-21 July in Salvador, Brazil. The Logistics League forms part of the Industrial League and is an application-driven league inspired […]