Blogroll

Paramount+ subscribers might be deeply invested in Yellowstone's sequel series, Dutton Ranch, which premieres tonight, or maybe even hunkered down with a South Park binge session. Either way, when the credits roll and the algorithm runs dry, finding your next watch can be a part-time job. I will be the first to recommend a good documentary to shake up your brain.

Family SUVs are changing fast, and buyers in 2026 are getting a lot pickier about what actually works day to day. Bigger engines and brute force used to be the default, but that thinking is starting to fade.

When the time comes to buy a portable external battery, you have to choose between the largest one you can afford or going for a smaller one. I'm here to argue that smaller is better.

The Outback has long stood apart from the sea of mainstream crossovers by offering something buyers could no longer find anywhere else: a rugged wagon that blended SUV practicality with a more car-like driving experience. For decades, that formula helped it become one of Subaru’s most recognizable and successful models. However, the redesigned 2026 version takes the nameplate in a very different direction, and not everyone is happy about it.

Phone hardware has been stagnating over the last few years. Unlike the 2010s, during which we saw gigantic steps forward with each phone generation, the 2020s are all about software, at least when it comes to Western phone brands, including Samsung.

Weeks after guru Mel Robbins received backlash for telling women to upload their banking statements to AI (specifically in an ad for Microsoft Copilot), OpenAI today announced a "personal finance experience" within ChatGPT.

The company is releasing a preview of the experience for Pro users in the United States, who can connect their financial accounts to the platform via the fintech software Plaid. Once their accounts are linked, ChatGPT can reason with actual monetary numbers and the user's shared priorities to help them spot patterns and plan for the future, OpenAI stated in its blog post introducing the product.

Users will be able to see a dashboard of their portfolio, spending, and upcoming payments. OpenAI's suggested use cases include financial goal planning, travel spend analysis, investment risk analysis, and subscription review.

Mashable 101 Fan Fave: Vote for your favorite creator today!

Credit: OpenAI SEE ALSO: OpenAI may sue Apple over ChatGPT integration

Pro users in the U.S. can start connecting their accounts today on a web browser or iOS, with support from thousands of financial institutions. OpenAI didn't list the institutions, but screenshots of the experience show American Express, Bank of America, Charles Schwab, and Robinhood as examples.

Integration with financial software company Intuit, which runs TurboTax, Credit Karma, and QuickBooks, is coming soon, OpenAI says.

In its announcement, OpenAI claims that 200 million people already use ChatGPT each month for budgeting, investment questions, and financial planning.

In this new financial mode, ChatGPT can't see full account numbers, but it can assess bank balances, transactions, and liabilities. Users can disconnect their financial accounts at any time, delete financial memories (used specifically for these conversations), or use temporary chats that won't appear in users' history.

How to get started with ChatGPT Finance Total Time

• 2 minutes

What You Need

• ChatGPT Pro in the U.S.

Step 1: Open Finances from ChatGPT's sidebar

Step 2: Select "Get started"

Step 3: OR, type "@Finances, connect my accounts" into ChatGPT

Step 4: Sync financial accounts with Plaid

Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.

Google has set steep requirements for Android 17's marquee Gemini Intelligence feature. They rule out most existing phones, to the point where some of the company's own Pixel 9 devices might not support the agentic AI.

Your DNS server may log every website you visit, and if you've never configured a privacy-respecting one, you're likely exposing your activity online. It's not a technical issue but a legal one. I'll go over which services I trust and why.

Emoji-obsessed Android users, get ready for Noto 3D emojis to take over your text messages, posts, and more. 

This week, Google announced a design overhaul to the nearly 4,000 unique emojis, which will soon be available on Android devices running the yet-to-be-released Android 17. According to Google, the previous emoji designs "often fall flat" when trying to express emotions online.

So, Google has introduced the new Noto 3D, its new line of 3D emojis. 

Mashable 101 Fan Fave: Vote for your favorite creator today!

"From our beloved innocent blob emoji of the 2010s to Noto 3D, the new emoji collection we announced today, emoji are the universal language of our digital lives and they’ve never felt more alive," Google said in a post.

The company also previewed what some of the 3D emojis look like in a short video animation, offering a comparison to the 2D designs that they're replacing.

Use the slider in the image to compare the 2D and 3D versions of the emoji:

Left: The new 2D emoji designs for Android. Credit: Google Right: The new 3D emoji designs. Credit: Google

The 3D emojis do have some pop to them, with more detail than the flat 2D designs that they are replacing.

According to Google, the new 3D emojis will first launch on Pixel phones later this year. Following the Pixel rollout, other Android devices will receive the 3D emoji update at a later date.

Want to learn more about getting the best out of your tech? Sign up for Mashable's Top Stories and Deals newsletters today.

Sony’s Project Ace robot has become the first to defeat an elite human table tennis player. Using nine cameras and real-time spin tracking, the system reacts with incredible speed and precision. Here’s how Sony is pushing robotics and AI-driven motion systems forward.

The pipeline feature is one of the driving forces behind the Linux philosophy, a single character that changes everything about how you work. By connecting the output of one command to the input of another, you can chain small programs together, creating a tool that is far greater than the sum of its parts.

Alienware just dropped the cheapest OLED gaming monitor we’ve seen, but are the savings worth it? Here's our opinion of the 27 240Hz QD-OLED.

For years, BMW’s sport sedans have been the benchmark if you wanted a car that could comfortably handle both your daily commute and a back-road blast without breaking a sweat. The M3 and M340i, in particular, built their reputations on delivering serious performance without completely sacrificing refinement.

Our recent paper, “LLMs Corrupt Your Documents When You Delegate”, has generated discussion about the reliability of AI systems in delegated workflows. We appreciate the interest in this work and want to clarify several important points about what the paper does—and does not—claim.

The research aims to develop robust evaluation methods for long-horizon delegated and collaborative tasks. More broadly, this work reflects an ongoing effort to better understand the gap between strong benchmark performance and certain real-world tasks. Using a controlled evaluation methodology, we examine how well information is preserved across these extended workflows. Within this constrained setting, we observe that models can accumulate fidelity degradation over repeated edits. Note however, that current production systems can mitigate these effects through verification loops, orchestration, and domain-specific tooling.

Our goal is not to argue against the use of AI systems in professional workflows, but rather to identify where current systems need further research and engineering to help make them more trustworthy collaborators. This benchmark is intended as a diagnostic tool for examining delegation patterns, not a measure of overall model capability, task success, or user outcomes.

Main results

The paper evaluates a specific interaction pattern we call delegated work—situations where a user entrusts an AI system to carry out multi-step modifications to important artifacts such as documents, spreadsheets, code, or structured files with limited human verification between steps.

We use chained transformation-and-inversion tasks that evaluate whether semantic content is preserved accurately across extended delegated workflows. Our evaluation uses domain-specific semantic parsing to focus on meaningful changes to the underlying artifact rather than superficial formatting or stylistic differences. The errors we report thus correspond to degradation in the underlying semantic content but, our measure of “corruption” did not include task completion or user satisfaction.

Using this methodology, we find that current frontier models can introduce sparse but consequential errors during long-horizon workflows, and that these errors may accumulate over repeated interactions. Across the evaluated settings, strong state-of-the-art models showed roughly a 19–34% degradation in artifact fidelity over 20 delegated iterations. Notably, Python workflows generally exhibited stronger robustness under extended delegated interactions, with less than 1% degradation on average.

Spotlight: Microsoft research newsletter

Microsoft Research Newsletter

Stay connected to the research community at Microsoft.

Subscribe today Opens in a new tab Methodological limitations

DELEGATE-52 was intentionally designed as a stress test for long-horizon delegated execution. The benchmark evaluates whether systems preserve artifact integrity across extended sequences of transformations and inversions.

The study focuses specifically on delegated execution with limited human intervention between steps. It does not attempt to measure the full range of real-world AI deployments, many of which involve substantially more oversight, verification, and workflow structure.

The paper also evaluated a simplified agentic harness with tool use capabilities such as Python execution and file operations. While this setup did not eliminate the observed degradation, it should not be interpreted as representative of production-grade systems optimized for specific workflows or enterprise domains.

Implications

We believe the primary implication of this work is that reliable long-horizon delegation remains an important open research and engineering challenge.

The results suggest that strong short-horizon benchmark performance alone may not guarantee dependable delegated execution over extended workflows. At the same time, the findings should not be interpreted as evidence that AI systems lack practical value in real-world work today.

In practice, many deployed AI systems combine models with specialized harnesses, orchestration layers, retrieval systems, verification procedures, memory mechanisms, and human oversight designed to improve reliability and deliver useful user outcomes despite underlying model limitations. We expect continued improvements in models, workflow-aware training, memory systems, and production-grade agentic harnesses to further reduce these failure modes over time.

Opens in a new tab

The post Further Notes on Our Recent Research on AI Delegation and Long-Horizon Reliability appeared first on Microsoft Research.

In a world that often thrives on being far too serious, outrageously funny TV shows deliver much-needed relief because they’re proudly unapologetic and completely unforgettable. They don’t just aim for a giggle — they go all-in with laugh-out-loud chaos, socially unacceptable characters, razor-sharp writing, and wildly entertaining moments.

The first smartphones had physical keyboards, and after over a decade of continuous slab phones, some of us still long to have our buttons back. Fortunately, the time for physical keyboards has returned, and these four brands are making it happen.

OnePlus appears to be scaling back its U.S. presence after withdrawing from other countries. Best Buy stores across the country have been pulling OnePlus phones from display shelves, limiting customers to online orders and some in-store pickups.

Certain Xfinity customers can now claim their portion of a $117.5 million class action lawsuit settled by communications giant Comcast.

Between Oct. 16 and Oct. 19, 2023, a third party accessed personally identifiable information for more than 35.8 million Xfinity customers, exposing usernames, passwords, contact information, and, in some instances, partial social security numbers. Multiple lawsuits against the company were consolidated into a class action case in 2024, alleging Comcast "failed to properly protect personal information" and had "inadequate data security" that led to the criminal cyberattack.

SEE ALSO: Meta AI Incognito Chat makes user conversations private, Zuckerberg says

The company officially settled the lawsuit in April, and a final approval hearing is scheduled for July 7.

Mashable 101 Fan Fave: Vote for your favorite creator today!

Settlement email: Is it real?

Comcast notified Xfinity customers impacted by the breach via email on December 18, 2023. Anyone who received an email notification at that time is eligible to file a claim (which means not all Xfinity customers can claim the cash).

Affected customers may have received additional notification emails from Kroll Settlement Administration LLC, the court-approved claims administrator. These emails, referencing both reference KSA and the Hasson v. Comcast Cable Communications case, will include individual class member IDs necessary to submit claims on the official settlement page.

If you've received notice via email or mail, but are unsure if its legitimate, call the settlement customer service line at (833) 319-2401 to confirm your eligibility.

How to claim your money

Settlement members have until September 14 to submit a claim online or via mail.

Individuals can receive a $50 fixed payout or request reimbursement for any "documented out-of-pocket losses and lost time, the cost of Identity Defense Services and Restoration Services, and court-approved attorneys’ fees and expenses" up to $10,000, according to the settlement website.

Members have until July 1 to exclude themselves or object to the settlement.

TL;DR: Live stream Chelsea vs. Man City in the FA Cup final for free on BBC iPlayer. Access this free live stream from anywhere in the world with ExpressVPN.

Man City are still fighting it out for the Premier League, but they've got another focus this weekend.

The FA Cup final takes place on May 16 at Wembley Stadium. Man City face off against Chelsea, a team that has really struggled for form this season. Chelsea are in danger of finishing in the bottom half of the Premier League standings, but they've managed to scrape through to this showpiece event. A win over Man City would at least end this campaign on a high.

If you want to watch Chelsea vs. Man City in the FA Cup final for free from anywhere in the world, we have all the information you need.

When is Chelsea vs. Man City?

Chelsea vs. Man City in the FA Cup final kicks off at 10 a.m. ET on May 16. This fixture takes place at Wembley Stadium.

How to watch Chelsea vs. Man City for free

Chelsea vs. Man City in the FA Cup final is available to live stream for free on BBC iPlayer.

BBC iPlayer is geo-restricted to the UK, but anyone can access this free streaming platform with a VPN. These handy tools can hide your real IP address (digital location) and connect you to a secure server in the UK, meaning you can unblock BBC iPlayer from anywhere in the world.

Live stream Chelsea vs. Man City for free by following these simple steps:

1. Subscribe to a streaming-friendly VPN (like ExpressVPN)

2. Download the app to your device of choice

3. Open up the app and connect to a server in the UK

4. Visit BBC iPlayer

5. Live stream Chelsea vs. Man City for free from anywhere in the world

Opens in a new window Credit: ExpressVPN ExpressVPN (1-Month Plan) $12.95 only at ExpressVPN (with money-back guarantee) Get Deal

The best VPNs for streaming are not free, but most do offer free-trials or money-back guarantees. By leveraging these offers, you can watch Chelsea vs. Man City without committing with your cash. This isn't a long-term solution, but it does give you enough time to live stream select fixtures from the FA Cup before recovering your investment.

What is the best VPN for BBC iPlayer?

ExpressVPN is the best choice for bypassing geo-restrictions to stream live sport on BBC iPlayer, for a number of reasons:

• Servers in 105 countries including the UK

• Easy-to-use app available on all major devices including iPhone, Android, Windows, Mac, and more

• Strict no-logging policy so your data is secure

• Fast connection speeds free from throttling

• Up to 10 simultaneous connections

• 30-day money-back guarantee

A two-year subscription to ExpressVPN is on sale for $68.40 and includes an extra four months for free — 81% off for a limited time. This plan includes a year of free unlimited cloud backup and a generous 30-day money-back guarantee. Alternatively, you can get a one-month plan for just $12.99 (with money-back guarantee).

Live stream Chelsea vs. Man City in the FA Cup final for free with ExpressVPN.

Is Apple's partnership with OpenAI about to fall apart?

According to both Bloomberg's Apple insider Mark Gurman and a new report in the New York Times, AI giant OpenAI is currently considering suing Apple over the companies' agreement to integrate ChatGPT into some Apple Intelligence AI features.

Apple recently reached a $250 million settlement in a class action suit over its failure to deliver on its AI promises. (Apple hasn't admitted to any wrongdoing.) Now, OpenAI may serve Apple with a breach of contract notice, the Times reports.

Mashable 101 Fan Fave: Vote for your favorite creator today!

In 2024, OpenAI and Apple surprised the industry by partnering together on Apple's upcoming AI features known as Apple Intelligence. OpenAI's ChatGPT was expected to be prominently featured in Apple products, such as the iPhone and MacBook, powering Apple Intelligence. 

In addition, OpenAI appeared to be a key technology behind Apple's relaunch of its voice assistant Siri as an AI assistant. 

According to the reports, however, OpenAI is unhappy with how the now two-year-old partnership has panned out so far. The fallout appears to stem from how Apple integrated ChatGPT into its products.

OpenAI reportedly expected that the partnership would encourage Apple users to subscribe to ChatGPT. Instead, OpenAI found that ChatGPT is difficult for users to even find within Apple Intelligence. On top of that, in January, Apple confirmed that it had struck a deal with Google to use Gemini to power Apple Intelligence, including Apple's long-awaited AI relaunch of its voice assistant Siri.

SEE ALSO: Apple picks Google Gemini over ChatGPT to power the new AI Siri

All of these issues have led OpenAI to consider taking legal action against Apple, claiming breach of contract.

We'll see if OpenAI actually follows through with a suit against Apple, or if the company is trying to renegotiate its agreement.

The AI company is currently awaiting the jury's decision regarding its own future following a trial stemming from the lawsuit Elon Musk filed against ChatGPT creator Sam Altman.

SEE ALSO: 'Memes on his phone!' Sam Altman's trial testimony takes a turn

Want to learn more about getting the best out of your tech? Sign up for Mashable's Top Stories and Deals newsletters today.