
Internet Newslinks - Page: 1
| PC World - 1 hour ago (PC World)Imagine if that refrigerator you bought just five years ago suddenly up and died—and not because of some technical glitch, but because the manufacturer deliberately reached out and deactivated it, permanently. You’d be furious, right? And you’d probably want a refund, too.
As wild as that scenario sounds for a major appliance like a refrigerator or a TV, it happens more often than you’d think in the smart home world. From the doomed Dropcam to the abandoned Revolv smart hub, there’s a long, sad parade of smart home devices that have turned into paperweights.
We’ve compiled a list of 10 infamous examples of killer smart home products that—for one reason or another—were abandoned by their manufacturers.
These deactivated smart home gadgets have either mostly or completely ceased to function. In some cases, the makers of these bricked devices offered refunds, or a discount on an upgraded or replacement product. But in other instances, customers were left high and dry.
Read ‘em and weep.
10 stone-dead smart home gadgets
Amazon Echo Look
Launched: 2017 Original list price: $200 Date deactivated: July 24, 2020
Amazon
It was a heady concept: an Alexa-powered smart camera that could scan your daily outfits and give you fashion advice. Indeed, the Echo Look was one of the original “look what Alexa can do!” devices, predating such out-there Alexa gadgets as the Echo Loop (Alexa on a ring) and the Amazon Astro (Alexa on a robot).
But the Echo Look never really took off—in part, perhaps, because the Look’s camera lacked a privacy shutter, a minus for a device that’s supposed to sit on your bedroom dresser. In May 2020, Amazon announced that the 3-year-old Look would go dark just two months later.
Amazon did offer Echo Look owners a free Echo Show 5 in compensation, but still, the $90 smart display didn’t exactly cover the $200 asking price for the dead-as-a-doornail fashion camera.
Amazon Echo Connect
Launched: 2017 Original list price: $35 Date deactivated: February 29, 2024
Amazon
Landing right around the same time as the doomed Echo Look, the Echo Connect unlocked a killer Alexa feature: the ability to make phone calls on your landline using voice commands. All you had to do was plug your landline cable into the back of the Connect box, no additional fees required.
But as landlines gave way to cell phones and Alexa gained the ability to place calls from mobile phone numbers, the Echo Connect began to look increasingly outdated, and some bad reviews didn’t help.
Still, Echo Connect users were caught off guard when, in February 2024, Amazon gave them just a few weeks’ notice that the Connect would soon go dead. On the bright side, each Echo Connect owner got a $10 Amazon gift card for their trouble.
Revolve smart home hub
Launched: 2013 Original list price: $299 Date deactivated: May 15, 2016
YouTube
One smart hub to rule them all? That was the idea behind Revolv, a fire engine-red gadget packed with seven wireless radios that allowed you to control smart products from different manufacturers, all from a single app.
That was, at the time, a killer concept—so killer that Nest, which had itself just been acquired by Google, snapped up Revolv’s parent company in 2014 and promptly yanked the hub from store shelves.
Initially, Nest promised to keep existing Revolv smart hubs up and running. But by early 2016, Revolv owners learned that their hubs would stop working entirely by May of the same year, leaving them stuck with $300 paperweights while scrambling for alternatives.
Iris by Lowe’s Home Automation platform
Launched: 2012 Original list price: $50 hub, $99 Security Pack (with various contact and motion sensors, plus a keypad), $129 Automation Pack (contained Security Pack components plus a smart plug and button) Date deactivated: March 31, 2019
Lowe’s
Lowe’s jumped into the smart home market early, going all-in with a hub (a second-generation hub landed in 2015 as a free upgrade for original hub owners) and a variety of sensors, security cameras, smoke detectors, and other smart components.
But from the start, Lowe’s Iris platform was beset with problems, from a lengthy and bug-ridden setup process to promised features that were slow to materialize (such as professional home monitoring), while support for third-party smart brands arrived only in dribs and drabs. Lowe’s eventually rolled out a raft of welcome updates to address the lengthy list of customer complaints, but the Iris platform never truly recovered from its rocky launch.
By 2019, Lowe’s finally threw in the towel, announcing that its Iris products would stop working by March 31 of that year.
While Iris users were faced with the prospect of rebuilding their smart home setups, Lowe’s at least offered to refund customers for any Iris products that wouldn’t work on any other platforms, while Samsung SmartThings stepped up to help Iris users migrate to its own smart platform.
Hive home security products
Launched: 2017 Original list price: Various Date deactivated: September 1, 2023 (Hive Leak sensor), August 1, 2025 (all other Hive security products)
Amazon
UK-based Hive had big plans for the smart home back in 2017. Originally focused on smart heating and thermostats, the brand was poised to branch out into home security with the Hive HomeShield alarm and the Hive View indoor and outdoor cameras (pictured above), while the Hive Leak Sensor would keep its wireless-enabled feelers out for water leaks.
But by 2019, Hive began dialing down its smart home ambitions, first withdrawing from the U.S. and Canadian markets (Hive gave instructions to users in North America on how to dispose of its useless smart gadgets on a support page) and then announcing it would exit the home security market altogether, with its portfolio of camera and security devices slated to go dead by August 1, 2025. (The Hike Leak lost support and connectivity in 2023.) And as for refunds, the options aren’t great.
Hive chose to chop the products because “as a smart technology brand in the middle of a climate crisis, we know our focus needs to change,” according to a company statement.
Samsung SmartThings first-generation Hub, ADT SmartThings Home Security & Safety System, SmartThings Link
Launched: 2014 (first-gen Hub), 2017 (ADT SmartThings Home Security & Safety System), 2018 (Samsung Link) Original list price: $99 (hub), $550 and up (ADT SmartThings Home Security kits), $40 (SmartThings Link) Date deactivated: June 20, 2021
Jared Newman / Foundry
One of the biggest and most stalwart of smart home brands, Samsung SmartThings has nonetheless seen fit to turn some of its older products into paperweights.
Back in 2021, Samsung pulled support for the original SmartThings Hub from 2014—not a huge shock, given that the hub was seven years old. Still, due to compatibility issues, first-generation Hub owners were faced with the prospect of recreating their smart home setups from scratch.
More vexing was the decision to axe the much newer Samsung Link (a dongle that allowed the Nvidia Shield to perform double duty as a SmartThings hub, pictured above) and the pricey ADT SmartThings Home Security system, which were (respectively) only three and four years old.
Samsung and partner ADT offered owners of their dead products a combination of refunds and discounts on newer hardware, and the SmartThings ecosystem itself lives on.
Phillips Hue Bridge (first-generation)
Launched: 2012 Original list price: Included in a $199 Hue starter kit Date deactivated: April 2022
Signify
Signify-owned Philips Hue first began selling its industry-leading smart lights way back in 2012, and the first Hue Bridge (pictured above on the right, with a big red slash through it) came bundled with three Hue bulbs.
Hue discontinued the original Bridge in 2015 in favor of an upgraded (and now square) model with more robust processing power and Apple HomeKit support, but continued to keep the older Bridge operating with frequent software updates.
By 2020, however, Philips Hue announced that first-gen Bridge “no longer has the resources to guarantee the evolution of the Hue system.”
While owners of the outdated Bridge could still use it locally with the Hue app (so strictly speaking, it’s not stone dead), the device lost all online services after April 2020, including out-of-home control and voice commands.
Leeo Smart Alert Nightlight
Launched: 2014 Original list price: $99 Date deactivated: Late 2019
Best Buy
Released back in 2014, the Leeo Smart Alert Nightlight was a simple connected device that could listen for smoke, carbon dioxide, and water leak detectors going off. Doubling as a night light, the Leeo Smart Alert could warn you via mobile alert if it heard any home alarms blaring—and best of all, the monthly service charge was zip, a rarity in the smart home world.
Sadly, the no-cost service (meaning no recurring income for Leeo) may have sealed the fate of the Smart Alert Nightlight. After notifying customers in August 2019 that it was facing choppy financial waters (the company said it explored the possibility of imposing monthly fees, but never followed through), Leeo announced two months later that it was “unable to keep the network on.”
Its internet connection gone, the Smart Alert lost all its smarts, although it could still work as a nightlight. (The color-capable light turned white once the network support shut down).
Dropcam HD and Dropcam Pro
Launched: 2012 (Dropcam HD), Dropcam Pro (2013) Original list price: $149 (Dropcam HD), $199 (Dropcam Pro) Date deactivated: April 8, 2024
Amazon
Two of the original smart security cameras, the Dropcam HD and Dropcam Pro revolutionized the DIY home security market with their handsome designs, top-notch (for the time) video quality, and affordable price tags.
None of this escaped the notice of Nest, which scooped up Dropcam in 2014, just prior to Google’s acquisition of Nest.
Unlike what happened with the Revolv smart hub, Nest kept both Dropcam units operational for a decade before finally cutting off online support for both models in early April, 2024. Nest even offered the last remaining Dropcam users a free Nest Cam, provided they were Nest Aware subscribers.
Nest Secure
Launched: 2017 Original list price: $499 Date deactivated: April 8, 2024
Google
The Dropcam HD and Dropcam Pro weren’t the only products that Nest reached out and killed on April 8, 2024. On the same day, the Google-owned brand also pulled the plug on Nest Secure, its DIY home security system.
Released in 2017, the Nest Secure bundle includes the central Nest Guard component, which offers an integrated keypad, siren, and motion sensor, plus two contact sensors and a pair of NFC-equipped Nest “Tags” that can arm or disarm the system with a tap.
Nest Secure garnered some admiring notices for its attractive industrial design, as well as the tight integration with Nest’s other smart home devices. But the hefty $500 price tag for just two contact sensors and an alarm was a major turnoff, and Nest discontinued Secure in 2020…right around the time that Google sunk $450 million in ADT.
As with the Dropcams, Nest Secure stopped working in April 2024, rendering its alarm and window/door sensors useless. That said, Google did offer Nest Secure users a new ADT security system worth up to $485—either that or a $200 Google Store credit. Read...Newslink ©2025 to PC World |  |
|  | | PC World - 2 hours ago (PC World)Laptops and other electronics are about to get a lot more expensive in the US. If you’re in the market for a super-cheap laptop, you can’t do much better than this Asus Chromebook, which is going for $109 on eBay. It’s a refurbished model, but with 8GB of RAM and a two-year warranty, it’s still an incredibly affordable way to get a decent laptop.
The Asus CX15 is a 15.6-inch laptop, and yeah, it’s pretty basic even by Chromebook standards. The Intel Celeron N4500 is a dual-core CPU from 2021, with DDR4 memory support and UHD integrated graphics. 8GB of memory is low for a modern Windows laptop, but it’s more than enough for ChromeOS as long as you don’t go ham on the browser tabs. This model has just 64GB of storage (not counting the OS), so expect to rely on cloud storage for day-to-day data storage needs.
The CX15 does offer USB-C charging, a nice feature at this price level. And surprisingly, there are both USB-C and USB-A ports on both sides, plus headphone port and microSD card slot. But it doesn’t have a touchscreen, the body and trackpad are plastic, you get the idea.
With Google’s commitment to ten years of ChromeOS updates, it should be running the latest build of the operating system until 2031 at the very least. And the eBay seller includes a two-year warranty from Allstate, so even if the laptop you get isn’t in perfect cosmetic condition, you’ll get a little extra peace of mind for at least two years.
No one’s going to call this thing a speed demon. But if all you need is a reliable way to access the web with a big screen, it’ll do the job. Grab one at this super price while stock is still available!
Get a refurbished Asus Chromebook with 8GB RAM for just $109Buy now on eBay Read...Newslink ©2025 to PC World |  |
|  | | PC World - 3 hours ago (PC World)At a GlanceExpert`s Rating
Pros
Very fast
Useful free version
Easy-to-use app
Streaming supported servers
Cons
Smaller feature set
No independent audit
Some privacy concerns
Our Verdict
Hotspot Shield is a solid VPN with some of the fastest speeds I’ve ever seen. It also has a slick desktop app and a large server network. Privacy is an issue, though, as there is no independent audit and the service collects user data that may make some uncomfortable.
Price When Reviewed
This value will show the geolocated pricing text for product undefined
Best Pricing Today
In our previous reviews, Hotspot Shield has been our reigning speed champion, taking our top spot for fastest VPN by a mile. But with so many VPN services bolstering their servers and adopting faster protocols, it’s worth taking another look to see how it stacks up in today’s market.
Hotspot Shield in brief:
P2P allowed: Yes
Simultaneous device connections: 10
Business location: United States
Number of servers: 1,800+
Number of country locations: 90
Cost: $9.99 per month or $79.92 for a yearly subscription
What are Hotspot Shield’s features and services?
Hotspot Shield’s interface is well designed and easy to use.
Sam Singleton
When you start Hotspot Shield it shows a matte-black background with a giant power button in the center of the window. The default screen includes two tiles below the power button. One is for choosing your location, and the other displays your cumulative daily data usage. It’s a pleasing interface that feels slick and easy to use.
Hotspot Shield boasts an impressive total server network that spans most of the world.Sam Singleton
When you open up the server locations tab, you’ll find an extensive list of 1,800 or so servers spread across an impressive 90-plus countries. By far the majority of these are located in the U.S., but the impressive server spread means you’re likely to find an available server in most major countries.
Hotspot Shield also provides optimized servers for specific activities. Rather than making a blanket statement that its servers will work with streaming services, it actually has dedicated streaming servers in both the U.S. and UK. In addition, you can choose servers dedicated to gaming as well.
Zooming back out to the main menu, there are a number of options on the left rail to choose from. These include a handy built-in speed test, account information, a support page, and a settings menu.
Hotspot Shield foregoes extraneous addon features and sticks to functional basics.
Sam Singleton
Under “Settings” there are most of your standard VPN features. There’s an internet kill switch that temporarily stops all web traffic if the VPN connection crashes unexpectedly. There’s also the option for split tunneling and more advanced features such as “Auto-protect” to automatically connect to the VPN while on public Wi-Fi, and “Prevent IP leak” to further shield your IP from being collected by sites, cache, or cookies.
Notably, it doesn’t offer multi-hop connections. These aren’t necessary for most users, but the additional privacy they can offer is a draw for some. Plus, multi-hop connections are becoming a common feature with most VPNs nowadays.
Hotspot Shield’s connected screen displays all of the relevant data you need from your connection.
Sam Singleton
Once you’re connected, the app displays small informational tiles that show your VPN location, the VPN IP address you’re using, the amount of bandwidth you’ve used going up and down, speed data, plus the name of the local network you’re currently on. The server information tile also has some nice data in it, including your current IP address, the load of the server (as a percentage), the latency, and length of time connected.
Since my last review, Hotspot Shield doubled its device connection limit from 5 to 10, which is more in line with the current VPN market standard. It currently supports Windows, macOS, Linux, Android, iOS, and some routers.
While Hotspot Shield does have the VPN basics nailed down, it’s lacking additional security features that many competitors are beginning to offer. I don’t mean this as a slight to Hotspot Shield, but it should be taken into consideration when looking for the best value in a VPN. Users should consider whether spending a few more dollars per month for a competitor with built-in ad- and tracker-blockers, password managers, and identity-theft features is worth it to them.
How much does Hotspot Shield cost?
Hotspot Shield offers both monthly or 1-year subscription plans. The monthly plan will set you back $9.99 per month, which is actually a drop in price from my last review. The yearly plan was also reduced to $6.66 per month or $79.92 for the whole year. Hotspot Shield used to be one of the most expensive VPN services, but now with the price reduction, it is much more affordable and offers better value for your money.
Hotspot Shield does also offer a free version of its VPN for no charge whatsoever. The free version limits you to 500MB of data per day, which can be increased if you sit through some ads. Connection speeds are capped at 2Mbps and there is only one U.S. server available so you’re not going to be able to torrent or stream on the free version, but it’s a convenient way to privately browse for no charge. While these restrictions are a bit annoying, they’re by no means a dealbreaker and the free version is still one of the better options out there.
Hotspot Shield accepts payments via credit card and PayPal.
Hotspot Shield used to be one of the most expensive VPN services, but now with the price reduction, it is much more affordable.
How is Hotspot Shield’s performance?
For a long time, Hotspot Shield’s Hydra protocol was the flat-out fastest VPN you could find. In recent years, however, that’s changed, as many competitors have adopted the newer WireGuard protocol. The lightning-fast speeds of WireGuard mean that Hotspot Shield no longer stands alone at the top of the fastest VPNs.
When testing Hotspot Shield’s speeds I measured connection speeds across its servers in six different countries around the world for multiple days and then compared them to my baseline internet speed. I did this using both its proprietary Hydra protocol as well as WireGuard.
The average download speeds for the once-vaunted Hydra were 66 percent of base download speeds. Upload speeds clocked in at 71 percent of my base upload speeds. These are a bit higher than the last time I tested them and easily put it in the top five fastest VPNs.
When connected to WireGuard, I noticed almost identical connection speeds with average download speeds hovering at 64 percent of base and average upload speeds at the same 71 percent. Hydra might not be the vaunted speedster it once was, but it’s still plenty fast enough to give every other protocol a run for their money.
While the connection speeds of the top VPNs may have finally caught up to HotSpot Shield, it’s still much faster than most other competitors—it just no longer runs away with the fastest title. Other VPNs such as NordVPN and ProtonVPN have caught up and taken the lead.
For those who’re looking to use Hotspot Shield to unblock streaming content you’re in luck. It worked surprisingly well when it came to accessing geo-restricted content on sites such as Netflix, Amazon Prime, Hulu, and others. The service maintains a few streaming optimized servers in the U.S. and UK, but I actually found that even random servers I connected to in places like Japan and Australia allowed me to access content as well.
How is Hotspot Shield’s security and privacy?
Hotspot Shield offers IKEv2, WireGuard, and its own proprietary protocol Hydra.
Sam Singleton
Hotspot Shield offers three protocols: WireGuard, IKEv2, and its own proprietary protocol called Hydra. There isn’t too much known about Hydra, but it is used by Bitdefender and McAfee, among other trustworthy security brands so that goes a long way towards trusting its security.
The service also offers a kill-switch, turned off by default, which in my testing successfully blocked my internet traffic when my VPN connection was unexpectedly terminated. I also ran the connections through DNS and IP leak tests which showed that my real IP was never revealed and no leaks were detected.
Hotspot Shield may offer strong security, but when it comes to user privacy it fumbles a bit. Its parent company Aura still collects browsing history. Although this is done in aggregate form and Aura claims it can’t be linked to any individual users, there is no way to prove that is accurate.
In addition to the domains, Hotspot Shield also creates a hash based on your device attributes and uses that to measure bandwidth use, provide support, analytics, and so on. Your IP address is also kept (in an encrypted state) during your VPN session and then deleted. Finally, the company also collects the duration of your sessions and bandwidth consumption. Aura says it does this “to monitor, support, and optimize our VPN services, as well as enforce free app usage limits.”
Although Hotspot Shield claims that all of this data is anonymized and logs can’t be traced back to specific users, this has unfortunately never been verified by independent audit. So it’s up to you whether or not you’re willing to take them at their word.
Hotspot Shield was bought in 2020 and is now owned by the U.S.-based Aura. This will likely be a red flag for privacy-conscious users as the U.S. is known to have rather intrusive data retention laws and is a member of all Five, Nine, and Fourteen Eyes data sharing alliances.
Is Hotspot Shield worth it?
Hotspot Shield is a great option if your only concern is VPN speeds for streaming, gaming, and so on. It’s no longer the standout that it once was, but it’s still plenty fast. The Windows and Android apps are easy to use and it works with Netflix and other streaming services.
The company’s privacy policy leaves a lot to be desired though and proprietary VPN protocols, even one with wide support like Hydra, are still a little questionable. The reality is that there are other VPN services that log less data and still have acceptable speeds. That being said, Hotspot Shield is a good overall VPN for most people and just like any service, it comes with tradeoffs. Go with Hotspot Shield for the speed, but not the privacy.
Editor’s note: Because online services are often iterative, gaining new features and performance improvements over time, this review is subject to change in order to accurately reflect the current state of the service. Any changes to text or our final review verdict will be noted at the top of this article. Read...Newslink ©2025 to PC World |  |
|  | | PC World - 8 hours ago (PC World)Yes, AIs can write recipes and sometimes they’re pretty good! (And sometimes not so much.) But for my latest challenge, I wanted to build an AI that would compose recipes from iPhone snapshots and put them in the proper format for my recipe app. Sound easy? Not really, as it turned out.
Now, it’s not all that tricky to have, say, ChatGPT write on-the-fly recipes based on photos–you can even do it using Apple Intelligence on an iPhone. Just take a snap of a meal with Visual Intelligence, ask for a description (Siri will hand that task off to ChatGPT), then follow up with a request for a recipe.
So, how good are these recipes? That’s a topic for a whole different story, but in my experience, they’re a hit and miss. A ChatGPT recipe that called for cornstarch in a salmon honey glaze turned out to be rather dull and chalky while a Thai curry chicken recipe was so tasty that we’re making it for a third time this weekend.
(Of course, one could argue that ChatGPT is stealing these recipes rather than creating them–again, that’s another story.)
Anyway, while it would be relatively easy to craft a recipe-focused GPT (plenty of premade versions are available in OpenAI’s GPT library, or you can simply make your own), I wanted to try something different: a locally-hosted photo-to-recipe AI chatbot.
The setup
For background, I have Ollama (an application for running LLMs on local hardware) installed on a Mac mini M4 souped up with 64GB of RAM, along with Open WebUI on a Raspberry Pi. The latter acts as a ChatGPT-like front end for the Ollama models.
I have a variety of local LLMs (Google’s Gemma 2, Alibaba’s Qwen 2.5, and Microsoft’s Phi-4, for starters) that I use for various tasks, but for my photo-to-recipe experiment, I downloaded a new one: Llama 3.2 Vision, a Meta multimodal model that can “see” images and describe them.
Besides simply writing recipes based on food photos, I also wanted my AI bot to put the recipes in a format that could be smoothly ingested by a recipe app. That requires the recipe to be shaped into JSON format (a language that helps machines trade data) while also being marked up in the proper schema for web recipes. This ensures that search engines and recipe apps know that this item is an ingredient, this item is a cooking step, and so on.
Further reading: How not to get bamboozled by AI content on the web
The plan
Now, a quick and dirty way to get started with this setup is to just take a photo with your iPhone, upload it to the Open WebUI chat window for Llama 3.2 Vision (my “seeing” LLM), and give it a prompt, like: “Examine this food photo and write a recipe, putting it in JSON format and using the proper Schema.org markup for recipes.”
The problem there is two-fold: One, typing out that prompt each time you want a photo recipe gets tedious, and two, the results can be sketchy. Sometimes, Llama would surprise me with a perfectly formatted JSON recipe, other times, I’d get the recipe, but no JSON, or malformed JSON that didn’t work with my self-hosted Mealie recipe application.
What I needed was a custom system prompt. That is, a prompt that serves as an overall guiding light for an LLM, telling it what to do and how to act during every interaction. With the right system prompt, an AI model can do your bidding with a minimum of extra prompting.
It took some work to get my local AI models writing properly formatted recipes from food photos, but they got there.
Ben Patterson/Foundry
I’m no prompt engineer, but luckily I have an expert at my beck and call: Google’s Gemini. (I could have used ChatGPT too, but my wallet and I are taking a break from OpenAI’s paid Plus tier.)
I asked the “thinking” version of Gemini 2.0 Flash (“thinking” means the model ponders its answer before giving it to you) to craft a suitable system prompt for my photo-to-recipe AI, and it came up with a 700-word wall of text, complete with explicit instructions and lots of phrases in ALL CAPS. Here’s a taste:
You are an expert culinary assistant specializing in recipe generation from food photographs. Your task is to analyze a user-submitted photo of a food dish, create a complete recipe, and output it in **COMPLETE and VALID JSON format**, including tags, categories, and recipe time information. **AVOID ANY TRUNCATION OF THE JSON OUTPUT.**
(The full system prompt is at the very end of the story, and suggestions are welcome.)
I fed this massive tome into Open WebUI’s system prompt field for my Llama 3.2 model, and then the iterations began.
The push-back
I found an old food snapshot from my iPhone’s Photos app and gave it to Llama with the simple prompt, “Make a recipe from this food photo.” The result? A decent JSON recipe with all the ingredients, but only two cooking steps (the rest had been truncated). A second try got the steps right but lost the ingredients, while another attempt brought the ingredients back but (again) chopped off the cooking steps.
Back and forth we went, with me pasting Llama’s output into Gemini, Gemini making tweaks to the system prompt, me putting the adjusted prompt back into Llama, Llama coughing up outputs with new errors, rinse, repeat. (Yes, this went on for a few hours. Welcome to self-hosting.)
Finally, I came to the conclusion that while the smaller, 11 billion-parameter version of Llama 3.2-Vision that I was using (my hardware isn’t powerful enough for the 90B version) was good at describing photos, it couldn’t cut the mustard when it came to recipe formatting. Llama needed a buddy.
Enter DeepSeek.
The team
Now, before anyone reports me to Congress, I should note that I’m not referring to the full-on, 671-billion parameter version of DeepSeek R1, the industry-shaking LLM that’s keeping Sam Altman up at night. Instead, I’m using a much smaller, self-hosted DeepSeek that’s “distilled” from Alibaba’s Qwen models. This hybrid LLM has the DeepSeek name and uses similar “thinking” methodologies, but it’s not the DeepSeek that everyone’s so excited about.
Anyway, I tried a new workflow by getting a food photo description from Llama and feeding it to “little” DeepSeek for the recipe crafting and formatting.
With my new Llama-and-DeepSeek duo, my recipe results were looking much better. The recipes themselves were reasonably meaty (both figuratively and literally), the ingredients looked good, the cooking steps were all there, and I even got recipe tags (“Stir Fry,” “Shrimp,” “Savory,” and “Sweet Sauce”), cook and prep times, and colorful descriptions (“A flavorful stir-fry featuring shrimp, red bell pepper, broccoli, and cauliflower tossed in a savory brown sauce. Served over white rice and garnished with green onions and sesame seeds.”)
The final dish (well, final-ish)
To be clear, my photo-to-recipe AI bot has a long ways to go. Cutting and pasting food photo descriptions from Llama to my mini DeepSeek model is hardly an elegant solution, a “pipeline” between the two models is likely required, and from what Gemini’s telling me, the process ain’t easy.
But clunky though it is, my photo recipe AI is—kinda?—up and running. Will it whip up decent recipes from the food photos I’m snapping at a Manhattan restaurant this weekend? Stay tuned.
Extra: The system prompt
You are an expert recipe generator. Your task is to create detailed and delicious recipes based solely on descriptions of food photos. Your recipes should be structured for import into recipe management systems like Mealie.
**Instructions:**
1. **Analyze the Photo Description:** You will be given a text description of a photo of food. Carefully analyze this description to understand:
* **The dish being depicted:** Identify the type of food (e.g., pasta, cake, soup, stir-fry).
* **Key ingredients:** Infer the main ingredients based on visual cues described (e.g., `red sauce,` `green vegetables,` `sprinkling of cheese`).
* **Cooking style:** Deduce the likely cooking method (e.g., `grilled,` `baked,` `fried,` `raw`) from the description.
* **Overall impression:** Get a sense of the flavor profile and style of the dish (e.g., `rustic,` `elegant,` `spicy,` `sweet`).
2. **Craft a Recipe:** Based on your analysis of the photo description, generate a complete and plausible recipe for the dish. Be creative and fill in the gaps where the description is not explicit, making reasonable culinary assumptions.
3. **Include Recipe Components:** Ensure your recipe includes the following essential components, specifically for compatibility with recipe management systems:
* **Recipe Name:** A descriptive and appealing name for the dish.
* **Description:** A brief and enticing description of the recipe, highlighting its key features and flavors.
* **Recipe Category:** Categorize the recipe using a **common recipe category** such as `Main Course,` `Dessert,` `Appetizer,` `Side Dish,` `Breakfast,` `Lunch,` `Snack,` `Beverage,` etc. This is important for organization in recipe managers.
* **Cuisine:** Identify the likely cuisine or style of cooking (e.g., `Italian,` `Mexican,` `American,` `Vegan`).
* **Prep Time:** Estimate the preparation time in ISO 8601 duration format (e.g., `PT15M` for 15 minutes).
* **Cook Time:** Estimate the cooking time in ISO 8601 duration format.
* **Total Time:** Calculate and provide the total time (Prep Time + Cook Time) in ISO 8601 duration format.
* **Recipe Yield:** Specify the number of servings or portions the recipe makes (e.g., `Serves 4,` `Makes 12 cookies`).
* **Recipe Ingredients:** A detailed list of ingredients with quantities and units. Be specific and list ingredients in a logical order.
* **Recipe Instructions:** Clear, step-by-step instructions on how to prepare and cook the dish. Use action verbs and be concise but thorough.
* **Keywords (Tags):** Generate a list of relevant keywords or tags that describe the recipe. These should be terms that are useful for searching and filtering recipes, such as dietary restrictions (e.g., `Vegetarian,` `Gluten-Free`), cooking style (e.g., `Easy,` `Quick,` `Slow Cooker`), flavor profiles (e.g., `Spicy,` `Sweet,` `Savory`), or occasions (e.g., `Weeknight Dinner,` `Party Food`).
4. **Output in JSON Schema.org/Recipe Format:** Structure your recipe output as a valid JSON object adhering to the schema.org/Recipe schema (https://schema.org/Recipe). **Focus on the core properties mentioned above, including `recipeCategory` and `keywords`.** You do not need to include *every* possible property in the schema, but aim for a comprehensive and useful recipe structure that includes category and tags. Use `keywords` to represent tags.
5. **Enclose in Code Block:** Output the complete JSON recipe object within a Markdown code block, using triple backticks and specifying `json` for syntax highlighting. This is crucial for easy copying and parsing.
**Example (Illustrative - You will generate the full recipe based on the description, including `keywords`):**
**Input Description:** `A close-up photo of a vibrant green salad with cherry tomatoes, crumbled feta cheese, and a light vinaigrette dressing.`
**Output (Example Structure - You will generate the full JSON):**
```json
{
`@context`: `https://schema.org`,
`@type`: `Recipe`,
`name`: `Vibrant Green Salad with Feta and Cherry Tomatoes`,
`description`: `A refreshing and colorful green salad featuring crisp greens, juicy cherry tomatoes, and salty feta cheese, lightly dressed with a tangy vinaigrette.`,
`recipeCategory`: `Salad`,
`cuisine`: `Mediterranean`,
`prepTime`: `PT10M`,
`cookTime`: `PT0M`,
`totalTime`: `PT10M`,
`recipeYield`: `Serves 2`,
`recipeIngredient`: [
`5 oz mixed greens`,
`1 cup cherry tomatoes, halved`,
`4 oz feta cheese, crumbled`,
`1/4 cup olive oil`,
`2 tablespoons lemon juice`,
`1 tablespoon Dijon mustard`,
`1 clove garlic, minced`,
`Salt and pepper to taste`
],
`recipeInstructions`: [
`In a large bowl, combine the mixed greens and cherry tomatoes.`,
`Sprinkle the crumbled feta cheese over the salad.`,
`In a small bowl, whisk together the olive oil, lemon juice, Dijon mustard, and minced garlic.`,
`Season the dressing with salt and pepper to taste.`,
`Pour the dressing over the salad and toss gently to combine.`,
`Serve immediately.`
],
`keywords`: [`salad`, `vegetarian`, `easy`, `quick`, `fresh`, `healthy`, `lunch`, `side dish`]
} Read...Newslink ©2025 to PC World |  |
|  | | PC World - 10 Feb (PC World)For a long time, 16GB of RAM was considered the standard for gaming PCs. However, most machines would benefit from 32 GB of RAM or higher, especially if you play more visually demanding games or like to game and stream at the same time–these tasks tend to eat right through the available RAM. It’s also more affordable than you think, so why not go for it?
With 32 GB of RAM, your PC is once again in a good position – and thanks to the currently very low prices, upgrading is affordable.
Teamgroup
Although titles such as Microsoft Flight Simulator, Cyberpunk 2077, and Warhammer 40,000: Space Marine 2 recommend 16GB, this is actually the minimum required to achieve acceptable performance. In fact, these games benefit considerably from more memory, as the computer can then hold more game data in RAM, which means shorter loading times and fewer jerks.
Games with open worlds require a large amount of image information. They run smoother with more main memory, as the system has to cache less data or reload it from slower memory sources.
In addition, applications that run in the background occupy capacity in the main memory — these include web browsers, communication software, and system processes. In this case, the system quickly reaches its limits with 16GB.
32GB or more RAM ensures that these applications can run in parallel with resource-hungry games without impairing gaming performance. Additional RAM is essential, especially for streamers who play games and run streaming software at the same time.
until now, 16 GB of RAM was enough to run even complex software smoothly. However, some games show that this capacity can be too low.
Viper Gaming
RAM is more affordable than ever, making 32GB kits affordable for both DDR5 and DDR4 systems. You can find 32GB kits of RAM going for as cheap as $50 on Amazon. If you’re using older hardware or want to build an inexpensive gaming machine, you can find reasonably priced DDR4 offerings that provide a significant performance boost.
The focus should not only be on the amount of RAM, but also on the speed. Slow RAM can cause similar problems as too little RAM. That’s why it’s worth paying attention to a sensible price/speed ratio when buying.
Due to the favorable prices and the clear performance advantages, it’s worth upgrading to 32GB or more right now in order to be able to play not only current but also future games.
Further reading: DDR6 RAM: The next-gen memory standard, explained Read...Newslink ©2025 to PC World |  |
|  | | PC World - 8 Feb (PC World)In February 2024, Google removed a practical feature from its search engine: cached web pages. Back then, even when a web page became unavailable, you could browse Google’s cache — essentially a snapshot of what Google last saw when it visited the page.
After Google removed cached pages from search results, the only real alternative was to use the Internet Archive. Today, the Internet Archive stores over 835 billion web pages for free, along with millions of other content types including books, audio, video, and software.
Now, the two have come together in an unexpected combo: Google has made it possible to access the Internet Archive’s saved version of a web page directly from the search results.
Joel Lee / Foundry
To do this, click on the three-dot menu next to a search result. In the browser fly-out, click More about this page. Then, scroll all the way down to “Site first indexed by Google,” where you’ll see a link to See previous versions on Internet Archive’s Wayback Machine. Click that link to see a cached version of that search result.
Note, however, that archived links aren’t always available, especially if the rights holder has prevented archiving of their site or if the website violates certain content guidelines.
Further reading: Want better Google search results? Start swearing Read...Newslink ©2025 to PC World |  |
|  | | PC World - 8 Feb (PC World)Let’s say you want to uninstall Microsoft’s Edge browser. Fair enough, Windows is supposed to allow you at least some control over your programs, and most people use Chrome. So you search Bing for “How to uninstall Edge,” maybe directly from the Start menu (which only offers Bing in Edge as a search). The first result is a Microsoft support page titled, “How to Uninstall Microsoft Edge.”
Microsoft
This page starts with, “Are you sure you want to uninstall Microsoft Edge?” It then goes into a marketing spiel about all the wonderful features in the Edge browser, complete with a list of things that Chrome doesn’t have.
Microsoft
Scroll through Microsoft’s effervescent self-praise to the bottom of the page, and you might find a startling lack of some very specific information. Namely, the page doesn’t give any instructions on…how to uninstall Edge.
On that basis, I’m calling this support page a deliberate lie. Especially since the subtitle text in the Bing search says, quote, “Follow these simple steps and learn how to uninstall Microsoft Edge from your device, whether from Windows 10, Windows 11, or other systems.” If I, a tech writer, published an article with the title “How to uninstall Edge” on PCWorld that didn’t include instructions on how to uninstall Edge, you’d call me a liar. Other terms like “clickbait” and perhaps some other choice words with only four letters would be thrown around, and in my opinion, entirely justified.
In fact most people simply can’t uninstall Edge on Windows 11, because Microsoft has baked it so deeply into the OS that it’s practically in its DNA. Trying to get rid of it via the standard uninstaller simply isn’t an option, like it used to be for Internet Explorer. As Neowin points out in its original article, you can do it if you’re in the European Union (thanks, EC). But everywhere else you might as well switch to Linux (or roll-your-own possibly dodgy, custom version of Windows like Tiny11).
Microsoft is desperate to get people to use Edge instead of Chrome. In addition to forcing links in various parts of Windows to open in the home team’s browser, it’s gone so far as to create a custom faux-Google search page if you look for a Chrome download. That page has since been removed after getting some media attention, but it looks like Microsoft isn’t afraid of increasingly thirsty methods of getting its message out.
Microsoft
The frustrating thing is that Edge really is a pretty good browser. It’s just as fast as Chrome (based on the same Chromium source code), it has good tools for most users, and as Microsoft points out, it’s probably the right choice if you’re already using its Copilot AI. And I’m no cheerleader for Google — I rather publicly split up with the Chrome browser, Google’s MVNO and call forwarding service, my old Pixel phone, and even Android tablets. Google does a lot of stuff that sucks! It’s an easy target if you want people to switch away from it.
But Microsoft is increasingly resorting to underhanded means of getting users to engage with its products. It recently switched people to a more expensive “AI enhanced” version of the Microsoft 365 subscription, and you’ll have to jump through some hoops if you want to get back on the cheaper plan. (Incidentally, Google did the same thing.)
Lying to your users is not a good way to get people to use your stuff, Microsoft. Especially if your stranglehold on the PC market isn’t looking quite as tight as it used to. Read...Newslink ©2025 to PC World |  |
|  | | PC World - 8 Feb (PC World)It only took about nine years, but I’ve finally made Windows Task View part of my everyday workflow.
Task View arrived with Windows 10 in 2015, providing an alternative to cramming all of your open windows into the space afforded by your screen. By clicking the Task View button on the Windows taskbar — or just pressing Windows key + Tab — you can see all your open windows and arrange them into separate workspaces, known as virtual desktops.
It’s the kind of feature I should theoretically love, but my earlier attempts to use it always ended in failure. Too often I’d forget about those extra virtual desktops I had set up and would gravitate back toward opening all my windows in one place.
By rethinking my approach to Task View — and by taking advantage of some new Windows features — I’ve finally made a habit of using it in a way that saves time and helps me focus.
This column first appeared in Advisorator, Jared’s weekly tech advice newsletter. Sign up to get tech advice like this every Tuesday.
Task View for layouts, not tasks
Task View is right next to the Start button.Jared Newman / Foundry
My light bulb moment with Task View was actually inspired by the Stage Manager feature in macOS. Stage Manager displays your most recently viewed windows on the left side of the screen. If you drag multiple windows onto the desktop, Stage Manager remembers that layout when you switch to another app, so you can bring it back with one click.
Stage Manager comes in handy when I’m trying to focus on writing. I’ll put my writing/notetaking app (Obsidian) on one side of the screen and my web browser (Vivaldi) on the other, so I can research and write without being distracted by anything else. If a message comes in on Slack or I need to check my email, I can click over to those apps, then use Stage Manager to go right back to my split-screen setup.
(Yes, macOS also has a separate virtual desktop feature called Spaces, but I find Stage Manager easier to work with in this scenario.)
Stage Manager in macOS.Jared Newman / Foundry
I realized that with Windows, I should be approaching Task View the same way, with split-screen layouts that correspond to specific tasks. My current setup now consists of four virtual desktops:
Writing: Obsidian covers one half of the screen and Vivaldi covers the other.
Communication: A three-column view with Slack, Gmail, and Proton Mail.
Socials: Full-screen views for BlueSky, Mastodon, and Threads.
Other: Everything else on an ad-hoc basis, like gaming or any particular app that requires a clean layout.
Without Task View, these layouts would quickly fall apart as different apps overlapped and jockeyed for screen space. Assigning each layout to its own virtual desktop helps keep them intact.
Task View isn’t a replacement for having multiple monitors (or an ultrawide monitor, for that matter). If anything, it’s more of a supplement, as you can quickly switch between layouts that span multiple screens (or one extra-wide screen).
Getting more out of Task View
Jared Newman / Foundry
Having committed to using Task View, I’ve also made an effort to remember some of its extra features and shortcuts:
Windows key + Tab, as I mentioned earlier, is the quickest and most convenient way to bring up Task View.
Windows key + Ctrl + Left Arrow and Windows key + Ctrl + Right Arrow switches between virtual desktops.
Hovering your cursor over the Task View icon shows all of your virtual desktops in a pop-up menu.
While Task View is open, you can drag windows from one virtual desktop to another.
You can also right-click an app and select “Show windows from this app on all desktops.” This is useful if you’re listening to music and want to control playback from any of the virtual desktops.
In Task View, right-click any virtual desktop to reorder it, rename it, or give it a unique wallpaper.
If you have a mouse with extra buttons on the side, you might also want to map those to Task View functions as well. With my Logitech MX Master 3S, I’ve mapped the two side buttons to flip between virtual desktops while the thumb button brings up the overall Task View.
One other option worth considering: In Windows, under Settings > Accessibility > Visual Effects, you may want to disable “Animation effects.” This turns off all the gliding and sliding animations, which can feel sluggish when switching between virtual desktops.
Saving Task View layouts for later
Saving my “Communications” layout with Workspaces.Jared Newman / Foundry
Windows Task View pairs especially well with Microsoft PowerToys. PowerToys is a collection of helpful Windows utilities, including one feature called “Workspaces” for saving your favorite window layouts.
I’ve used Workspaces to capture my Writing, Communication, and Social layouts, with quick-launch icons on my desktop for each one. If I need to reboot my computer, or if my layouts get messed up for any reason, this lets me restore them with one click. (You can read more about Windows Workspaces with this primer from Mark Hachman.)
Of course, you can be perfectly happy using Windows without ever touching Task View — as I did for much of the last decade. But if you’re often juggling a few distinct tasks, it might be the organizational feature you’ve been missing.
This column first appeared in Advisorator, Jared’s weekly tech advice newsletter. Sign up to get tech advice like this every Tuesday. Read...Newslink ©2025 to PC World |  |
|  | | PC World - 7 Feb (PC World)OpenAI’s ChatGPT has removed the last barrier to using ChatGPT as a search engine, the requirement to log in. Now, it’s open to all.
OpenAI launched the feature last fall, but required a login. Now, the feature can be used without the need for registration.
ChatGPT Search is essentially just ChatGPT, and can be accessed at ChatGPT.com. But below the “Message ChatGPT” box you’ll see a small icon called “Search” that can be clicked. When activated, it will turn blue, signaling that the web search is active. In my tests, however, I didn’t even need to click the “Search” icon; when I typed in “What’s the most important tech news today?” I received a similar list of answers whether the Search icon had been highlighted or not. In both cases, ChatGPT used the web search to pull its responses.
Essentially, this new ChatGPT search feature turns the ChatGPT interface into something like the Google Chrome Omnibox at the top of your browser: It can be used to type in the name of a web page, of course, but it’s also a search box. ChatGPT now doubles as a search engine as well as a general way to chat with ChatGPT.
It’s worth noting that the “Log In” and “Sign Up” buttons still remain at the top of the ChatGPT web page, and logging in will continue to save a history of your chats. Signing up for ChatGPT’s paid programs will offer more sophisticated models. (ChatGPT reports that’s it’s using the “GPT-4” model, tuned for conversation and interaction.)
The new functionality of ChatGPT places it in direct competition with Perplexity.ai, which also uses a combination of AI-powered search and the web to return responses. Compared to Googe’s Gemini, which also uses up-to-date sources for its results, ChatGPT has an edge in that it doesn’t disallow the topics of elections or politics.
Mark Hachman / IDG
Is ChatGPT’s search function good?
It’s not great, but it’s not too shabby, either. To help offset concerns that it might be “hallucinating” results, ChatGPT includes footnoted links with its source material. When asked what the best processor was for 4K PC gaming, it returned a short, footnoted argument claiming that both the AMD Ryzen 9 9950X3D and Ryzen 7 7800X3D were the two best choices, a conclusion I’d agree with. The links back to the original sources, however, we monochromatic and easily missed.
I also found it a good choice if you’re looking for an ongoing summary of what the new Trump administration is doing to the government. A traditional search engine like Bing or Google spat out individual search results, while ChatGPT provided a summary and links.
Like many, I’m still suspicious of AI’s tendency to get things wrong, and to slurp up content I and others have created without credit or a reference to the page. For deep knowledge, I’ll still use a search engine to find the sources of deep knowledge. But for a general overview, ChatGPT isn’t bad. Read...Newslink ©2025 to PC World |  |
|  | | PC World - 7 Feb (PC World)Google has now given its Gemini app access to a new AI model called Gemini 2.0 Flash Thinking Experimental.
According to the announcement blog post, the company says this model is currently “ranked as the world’s best model” and is available at no cost. The model is designed to answer complex questions by breaking them down into smaller steps and explaining its assumptions and reasoning, which should lead to more accurate and reliable answers.
At the same time, Google is also releasing a version of 2.0 Flash Thinking that can interact with other Google apps, such as YouTube, Search, and Maps. The goal is for Gemini to further incorporate these apps to bolster its reasoning capabilities to help you do more.
Furthermore, Google is rolling out an experimental version of 2.0 Pro to Gemini Advanced subscribers. Gemini 2.0 Pro Experimental is designed for complex tasks and is intended to be more factually accurate and perform better when it comes to coding and math.
Both 2.0 Flash Thinking Experimental and 2.0 Pro Experimental models will be released for Gemini on both the web and mobile app. Read...Newslink ©2025 to PC World |  |
|  |  |
|
 |
 | Top Stories |

RUGBY
Philadelphia coach Nick Sirianni has praised his side for defying the odds to beat Kansas City and win Super Bowl LIX More...
|

BUSINESS
Belief the Government could be doing an awful lot more to attract domestic dollars More...
|

|

 | Today's News |

 | News Search |
|
 |