Feed aggregator

• Developers shift from loyalty to flexibility as OpenAI leads, but DeepSeek gains ground fast
• Microsoft struggles for developer mindshare with only a 10% adoption rate
• AI integration gets leaner as teams drop formal structures and focus on usable tools

OpenAI has been dominating the LLM space for some time, and remains the leading choice for many developers, but market dynamics and loyalty seem to be shifting, new research has claimed.

A survey by Vercel of 656 app builders found 87% use OpenAI’s models and 83% rely on its inference APIs. However, developers now use an average of two providers, and 60% have changed vendors within the last six months - raising questions about their loyalty.

DeepSeek is now used by nearly a third of businesses (29%), while Microsoft's LLMs, by contrast, appears to be struggling, with only 10% of respondents report using the latter, and 9 out of 10 developers do not consider it a viable option.

Developers now focus on tools, not teams

“AI is dissolving the boundaries between roles. We’re seeing new product designers blend UX, UI, and code in one creative flow - thanks to tools like Vercel, v0, Uizard, and Cursor. Whether junior or senior leader, anyone can now build, test, and ship ideas independently - and that’s not just efficient, it’s liberating,” said Nicolas Le Pallec, CTO, EMEA - AKQA.

AI now builds around clear use cases, with developers prioritizing tools - not teams. Forty-five percent of respondents said they have no dedicated AI team, while 57% reported no specific AI leadership structure. Instead, success with AI depends on clear priorities and the right tools.

“By embracing cutting-edge AI technologies, we’re empowering our teams to work smarter and faster,” said Dr. Jan Ittner of BCG X, echoing the sentiment that an AI writer or developer tool can be more valuable than another hire.

Product AI features are now a priority for 75% of customer-facing apps, while only 39% still include traditional support chatbots. Website personalization remains underutilized at 27%, indicating room for future innovation.

Cost control is also a major priority. Over 70% of developers manually test their models, spending under $1,000 monthly - showing how much can be achieved on limited budgets. To cut costs, only 14% train their models, while 60% instead use retrieval-augmented generation (RAG) and vector databases.

The focus on tools stems from a perceived "overhype" around AI, with an average rating of 6.4 out of 10. Yet developers gave a 7.7 out of 10 when asked if AI will transform their industries in the next 12 months.

In a shifting field where speed, precision, and adaptability matter more than scale, the best LLM for coding isn’t fixed - it’s the one that solves today’s problems while keeping tomorrow’s door open.

You may also like

• Check out the best AI phones on the market
• Here is our list of the best AI website builders on the web
• AI in the workplace: why upskilling, not fear, is the key to AI collaboration

The shipping company's temporary new policy comes as President Trump continues to upend U.S. global trade policy.

(Image credit: Nicolas Guyonnet/Hans Lucas)

Here are hints and answers for the NYT Strands puzzle No. 415 for April 22.

Here are hints and the answer for today's Wordle No. 1,403 for April 22.

Hints and answers for Connections for April 22, #681.

Don't worry if you miss the peak -- the Lyrids will continue through Saturday.

With award-winning tech and ultimate cleaning performance, Dreame's new line of robotic pool cleaners let you keep your pool sparkling without breaking a sweat.

• Synology’s 2025 Plus range only works with certain hard drives
• It says this is for lower failure rates and compatibility issues
• Older models won’t be affected by this new policy

Taiwanese NAS-maker Synology has confirmed a new policy that requires Synology-branded hard drives to be used for full compatibility within its 2025 Plus series of network-attached storage (NAS) devices.

From launch, the policy states Synology-branded hard drives will need to be used, but the company promises to “update the Product Compatibility List as additional drives can be thoroughly vetted in Synology systems,” a company spokesperson said(via Ars Technica).

In the meantime, this could spell out reduced choice and increased cost for customers, with potential sustainability issues stemming from users’ inability to use hard drives they might already have, rendering them unusable.

Synology NAS devices hard drives

Synology has nine products in its Plus range, supporting all the way up to 24 hard drives, however users will be required to use approved storage drives only.

The company said that “Synology systems are at less risk of drive failure and ongoing compatibility issues.”

Chad Chiang, the Managing Director of Synology’s Germany and UK businesses, noted: “With our proprietary hard disk solution, we have already seen significant benefits for our customers in various deployment scenarios.”

Despite the change for 2025-onward models, the company confirmed that older models sold up to and including 2024 (excluding XS Plus series and rack models) remain unchanged.

Interestingly, Synology doesn’t manufacture its own hard drives, instead choosing to rebrand models from popular brands like Toshiba and Seagate.

This could be good news, indicating the possibility that further Toshiba and Seagate models might appear on the approved list of third-party drives at a later date, however it could also mean that drives from other brands might be less likely to appear on the list.

Of course, this is speculation only, but we’ve contacted Synology for further details in the hope that we can get access to a list of pre-approved third-party drives.

You might also like

• We’ve listed the best NAS hard drives
• Keep your data online with the best cloud storage solutions
• Popular NAS vendor issues emergency patch against potentially highly damaging vulnerabilities — here's what you need to know

Born Jorge Mario Bergoglio in Buenos Aires, Argentina, Pope Francis died Monday, the day after Easter, at the age of 88. As the leader of the Roman Catholic Church, he was the first Latin American pontiff, the first non-European pope in more than a millennium, and he left behind a legacy of service, unity, and compassion.

(Image credit: Lisa Maree Williams)

You can even get store credit in some cases when you trade in old PCs, laptops and printers.

Looks like Bethesda really has been reworking Skyrim's predecessor for modern audiences.

Bluesky, the upstart, decentralized social network that's offered an island of blue calm for those departing the tumultuous, dark waters of X (formerly Twitter), is finally introducing an official form of account verification. It will look familiar to those who spent time on old-school Twitter, but it does feature a few notable caveats.

Up until today (April 21, 2025), there was no Bluesky-supported system for verifying an account or visual indication of verified status, and the microblogging social media platform was suffering from an imposter problem. Bluesky addressed this, in a way, by encouraging people to set up and use their own domain names to establish the veracity of an account.

My account, for instance, is attached to lanceulanoff.com, a domain I've owned for years. Bluesky reports that some 270,000 accounts have linked their accounts to domains.

Still, setting up a domain and connecting it to your Bluesky account is not a trivial matter. This new feature significantly simplifies the process.

A new check

(Image credit: Future)

According to a new Bluesky blog post, there are now three levels of identity on the platform: the basic Bluesky account, a Trusted Verifier, and a Verified Account.

The Trusted Verifier is interesting because it's a verified account that, with Bluesky's review, can verify other accounts. The example given is The New York Times' Bluesky account, which can then verify its journalists' accounts.

Years ago, Twitter had something similar, where an entity like TechRadar could ask Twitter directly to verify some of its journalist employee accounts.

It was not a popular feature among, for instance, celebrities and officials, who wondered (often on Twitter) why journalists needed verification. The topic of Blue Checks and verification on Twitter became so fraught that, for a time, Twitter's then-CEO Jack Dorsey stopped verifying accounts. When Elon Musk took over, he removed verifications from millions of accounts, only to return them some months later.

Who verifies who

Bluesky wrote that it is proactively verifying some "authentic and notable accounts," which will now have a white check in a blue circle.

Trusted Verifiers will have a scalloped blue check to indicate their vaunted status. A tap on someone's verified status can show you which Trusted Verifier granted verification.

It's not necessarily a foolproof system, as I think media companies might seek to verify all journalists in their camp, and Bluesky could push back against that. Or maybe Bluesky says yes, but at some point, "the notables" grow irritated because there are suddenly so many non-famous people with checks.

One thing Bluesky is not supporting is "requests for verification." However, the social media platform is not ruling it out and has promised that as this system settles in and stabilizes, it will "launch a request form for notable and authentic accounts interested in becoming verified or becoming trusted verifiers."

If you're wondering about the other decentralized social media platform, Threads, it adopts verifications directly from Instagram, another Meta property.

It'll be interesting to see what Bluesky asks for in its future verification process and if any part of it will involve some form of ID.

You might also like

• What is Bluesky? The new social media network explained
• How to switch to Bluesky from Twitter
• I wasted my day on Bluesky Social and no, I'm not sorry

• The FBI is warning about an ongoing scheme targeting victims of online fraud
• The victims are encouraged to reach out to a person on Telegram, posing as the chief of IC3
• The person would try to gain access to the victims' financial accounts

Cybercriminals are preying on victims of online fraud, using their state of emotional distress to cause even more harm, the FBI has said, revealing it received more than a hundred reports of such attacks in the last two years.

In the campaign, cybercriminals would create fake social media profiles and join groups with other victims of online fraud. They would then claim to have recovered their money with the help of the FBI's Internet Complaint Center (IC3). This makes the ruse credible, since IC3 is an actual division of the FBI and serves as a central hub for reporting cybercrime.

Those who believe the claim are then advised to contact a person named Jaime Quin on Telegram. This person, claiming to be the Chief Director of IC3, is actually just part of the scheme. Quin will tell people who reach out that he recovered their funds and would then ask for access to their financial information, to steal even more money.

Get Keeper Personal for just $1.67/month, Keeper Family for just $3.54/month, and Keeper Business for just $7/month

​Keeper is a cybersecurity platform primarily known for its password manager and digital vault, designed to help individuals, families, and businesses securely store and manage passwords, sensitive files, and other private data.

It uses zero-knowledge encryption and offers features like two-factor authentication, dark web monitoring, secure file storage, and breach alerts to protect against cyber threats.

Preferred partner (What does this mean?)View Deal

Getting in touch

This is just one example of how the scam works. The FBI says that initial contact from the scammers can vary.

“Some individuals received an email or a phone call, while others were approached via social media or forums," it said. "Almost all complainants indicated the scammers claimed to have recovered the victim's lost funds or offered to assist in recovering funds. However, the claim is a ruse to revictimize those who have already lost money to scams."

To minimize the risk of falling victim to these scams, you should only reach out to law enforcement through official channels. Furthermore, you should keep in mind that law enforcement (especially those in executive positions) will never reach out to you this way, especially not to initiate contact.

Finally, the police will never ask for your password, financial information, or access to private services.

Via BleepingComputer

You might also like

• Key Strategies for financial institutions to combat fraud
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

Pope Francis, one of the most popular popes in decades, has died. As the world mourns and reflects on his legacy, NPR would like to hear your stories.

(Image credit: Spencer Platt)

The pope's death at the age of 88 ends a 12-year papacy that began with the historic resignation of Francis' predecessor in 2013.

(Image credit: Imtiyaz Khan)

• Fake PDF converters are tricking users with cloned sites and fake CAPTCHAs
• PowerShell command installs malware that steals browser and crypto wallet data
• Attackers use realistic designs and social engineering to avoid detection

Cybercriminals are using fake PDF converters to install powerful malware on victims' systems, experts have warned.

Research from CloudSEK found attackers are cloning popular file conversion websites like pdfcandy.com - replicating its logo and brand elements - in order to trick users into downloading malicious software.

CloudSEK says these fake sites look almost identical to the real ones. When someone tries to convert a file, the page shows a fake loading screen and then prompts for a CAPTCHA verification. Instead of just confirming the user is human, this step leads to an instruction to run a PowerShell command. Following the command downloads a zip file containing malware known as ArechClient2, part of the SectopRAT family of information stealers.

Collecting personal data, and worse

The malware uses a number of hidden methods to infect the system. It spawns normal Windows processes to hide its activity and begins collecting browser passwords, crypto wallet information, and other sensitive data. Once the malware is active, it can quietly send stolen information back to the attackers, CloudSEK reports.

The FBI has already warned that online file converters are becoming a popular way for criminals to spread their malware. CloudSEK’s research shows that attackers are improving their methods, cleverly blending realistic website designs with social engineering tricks in order to lower users' defenses.

With online tools becoming part of everyday work and personal life, it’s important to know how to avoid these threats.

How to stay safe

(Image credit: Amazon India)

The best way to protect yourself is to avoid clicking random search results for online file converters. Always visit known official websites directly.

In addition to that, always double-check the website address for small spelling changes that might be easy to miss.

For a good starting point, check out our round up of the best PDF editors, and the best free PDF editors. We also recommend the best Adobe Acrobat alternatives.

Staying cautious when uploading documents online can stop many of these attacks before they start.

Keep your antivirus software up to date (you’re doing this anyway, right?) and scan any downloaded files before you open them. Installing browser extensions that block suspicious or dangerous sites can also help.

If a website asks you to run PowerShell commands or download extra files after uploading a document, close the page immediately.

Finally, if you think you’ve been tricked, disconnect the device from the internet right away, change all important passwords from a safe device, and let your bank or service providers know as soon as possible.

You might also like

• Stay protected with the best endpoint protection tools around
• We've also rounded up the best ransomware protection tools
• And these are the best free PDF to Word tools you can use

The New York Times reported that he shared information last month to a second private Signal group chat about upcoming strikes in Yemen.'/>

The White House has begun the process of looking for a new secretary of defense, according to a U.S. official who was not authorized to speak publicly.

(Image credit: Chip Somodevilla)

The last pope to die during his tenure was Pope John Paul II in 2005. Here's what happened in the days and weeks that followed.

(Image credit: Joe Raedle)

While Verizon Fios stands out as Harrisburg's top internet choice, it's worth exploring all your options before committing to a provider.

The best AI generation trends are the cute ones, especially those that transform us into our favorite characters or at least facsimiles of them. ChatGPT 4o's ability to generate realistic-looking memes and figures is now almost unmatched, and it's hard to ignore fresh trends and miss out on all the fun. The latest one is based on a popular set of Anime-style toys called Chibi figures.

Chibi, which is Japanese slang for small or short, describes tiny, pocketable figures with exaggerated features like compact bodies, big heads, and large eyes. They are adorable and quite popular online. Think of them as tiny cousins of Funko Pop!.

Real Chibi figures can run you anywhere from $9.99 to well over $100. Or, you can create one in ChatGPT.

What's interesting about this prompt is that it relies heavily on the source image and doesn't force you to provide additional context. The goal is a realistic Chibi character that resembles the original photo, and to have it appear inside a plastic capsule.

The prompt describes that container as a "Gashapon," which is what they're called when they come from a Bandai vending machine. Bandai did not invent this kind of capsule, of course. Tiny toys in little plastic containers that open up into two halves have been on sale in coin-operated vending machines for over 50 years.

If you want to create a Chibi figure, you just need a decent photo of yourself or someone else. It should be clear, sharp, in color, and at least show their whole face. The effect will be better if it also shows part of their outfit.

Here's the prompt I used in ChatGPT Plus 4o:

Generate a portrait-oriented image of a realistic, full-glass gashapon capsule being held between two fingers.

Inside the capsule is a Chibi-style, full-figure miniature version of the person in the uploaded photo.

The Chibi figure should:

• Closely resemble the person in the photo (face, hairstyle, etc.)
• Wear the same outfit as seen in the uploaded photo
• Be in a pose inspired by the chosen theme

A time capsule

(Image credit: Mr Rogers Chibi generated by ChatGPT)

Since there's no recognizable background or accessories in the final ChatGPT Chibi figure image, the final result is all about how the character looks and dresses.

I made a few characters. One based on a photo of me, another based on an image of Brad Pitt, and, finally, one based on one of my heroes, Mr. Rogers.

These Chibi figures would do well on the Crunchyroll Mini and Chibi store, but I must admit that they lean heavily on cuteness and not so much on verisimilitude.

Even though none of them look quite like the source, the Mr. Rogers one is my favorite.

Remember that AI image generation is not without cost. First, you are uploading your photo to OpenAI's server, and there's no guarantee that the system is not learning from it and using it to train future models.

AI image generation also consumes electricity on the server side to build models and to resolve prompts. Perhaps you can commit to planting a tree or two after you've generated a half dozen or more Chibi AI figures.

You might also like

• The action figure trend is the latest way people are ...
• Everyone wants the viral AI doll – but it's a privacy ...
• The hottest new AI image trend is Sora users begging for ...