Welcome back to The Prompt,Chinese AI company DeepSeek has taken the AI world by storm. News of the models capabilities briefly shaved $600 billion of market cap from chip giant Nvidia, which then jumped back and recovered almost half of it. With its latest R-1 model (free and open source) on par with OpenAIs o1 ($200 per month and locked away for paying users), AI-based application startups like Perplexity have already started using the model in their products. Elsewhere companies are rethinking their AI training strategies, multiple founders told me this week.Though the model received accolades from prominent AI leaders like OpenAI CEO Sam Altman, some in the space remain skeptical that DeepSeek was seemingly able to train its models with fewer resources, thus making it cheaper. OpenAIs Chief Research Officer Mark Chen said on X: I think the external response has been somewhat overblown, especially in narratives around cost.The model is also riddled with safety risks, according to a report by AI evaluation company Chatterbox Labs, and can produce hate speech, misinformation and sexually explicit content. Researchers at cyber intelligence company Kela were able to bypass its safeguards to use the model to create malware, as well.Now lets get into the headlines.SHOW ME THE MONEYPresident Donald Trump's announcement of $500 billion AI infrastructure project Stargate, jointly funded by OpenAI, Oracle, Japan-based SoftBank and Emirati-based investment firm MGX, resulted in major drama with tech leaders criticizing the megadeal. Elon Musk called the initiative fake, alleging that the companies involved dont actually have the money. He also called Sam Altman, a swindler. Microsoft CEO Satya Nadella commented: All I know is Im good for my $80 billion, allotted towards putting up data centers. Not wanting to be left behind, Meta followed up the Stargate announcement by saying it plans to plunge $60 billion to better its AI models.HUMANS OF AIA new startup called Palona launched by a team of former Meta and Google researchers is building an AI sales agent that simulates emotional intelligence. CEO Maria Zhang said the company uses one AI model to supervise another and keep track of the conversation. She claims this also ensures the AI system is accurate but is also persuasive, humorous and has personality. The company has raised $10 million in seed funding.AI DEAL OF THE WEEKClay, which uses artificial intelligence to filter sales leads, has raised $40 million at a $1.3 billion valuation. Among its customer base of 5,000 companies are outfits like Anthropic and OpenAI, which uses Clay to test sales tactics by surfacing helpful data points.DEEP DIVEThe power of DeepSeeks model and its pricing are already shifting the way American AI startups run their businesses. Its a cheap, compelling alternative to offerings from incumbents like OpenAI, Jesse Zhang, CEO of Decagon, which builds AI agents for customer service, told Forbes. DeepSeeks new model will likely force American AI giants like OpenAI and Anthropic to reevaluate their own prices.Eiso Kant, CTO and co-founder of Poolside AI, a unicorn that builds AI for software engineering, told Forbes that DeepSeeks strength is in its engineering ability to do more with less.What DeepSeek is showing the world is that when you put a strong emphasis on making your training compute-efficient, you can do a lot, he said. There's incredible things that you can continue to squeeze out of these Nvidia chips to make them incredibly more efficient.Others are less impressed. Writer CEO May Habib told Forbes shes not surprised that DeepSeeks models, trained on a significantly smaller budget, are able to match the most intelligent models in the US. In October, Writer launched a model that was trained with just $700,000, when it cost $4.6 million for OpenAI to build a model with similar capabilities. The company used synthetic data to lower its training costs.Even before DeepSeeks model exploded on the scene, we have been saying that these models are commoditizing. Theyre getting more and more distributed, Habib said.Read the full story on Forbes. AI INDEX Turing, a Palo-Alto-based company that provides data sourced from millions of human experts to train leading AI models from companies like OpenAI, Anthropic and Meta, announced today that it became profitable in 2024. CEO Jonathan Siddharth told me hes building an AGI infrastructure company, providing data to make models more intelligent and perform advanced tasks like coding and reason with complicated prompts. Turing, valued $1.1 billion in 2021, also assesses AI models performance and builds AI applications across sectors like healthcare and retail.$300 million Annual recurring revenue that Turing recorded in 2024, growing 3x from 2023.4 millionCoders, scientists and professionals contribute to what Siddharth claims is the worlds largest human intelligence cloud.QUIZ This company is using AI to better detect and remove AI-generated child sexual abuse material from the internet.PerplexityGenpactHiveMetaCheck if you got it right here.MODEL BEHAVIORWhen Google introduced AI-generated summaries that sat on top of links, the immediate result was hallucinated inaccurate responses that told people to eat rocks and add glue to pizza. But a longer-drawn, lesser-known ripple effect is now playing out: a reduction in traffic to individual websites, which is beginning to hurt companies like Kayak, Yelp and TripAdvisor, Forbes

Reviewers LikedSuper small form factorGreat 10Gbps performance1-, 2- or 4TB capacitiesHandsome designDrop-proof up to 7.5 feet256-bit AES hardware-based encryptionFive-year warrantyReviewers Didn't LikeNot the least expensive 10Gbps SSD availableLacks USB-C-to-A cable or adapterCompetitors and Related Products Our editors hand-pick related products using a variety of criteria: direct competitors targeting the same market segment, or devices that are similar in size, performance, or feature sets. Expert reviews and ratings 80 The X9 Pro is one of the fastest SSDs around, and its diminutive size combined with rugged design makes it a great choice to use on the road. The 2TB and 4TB models are a better value than their immediate rivals, and you get transfers going up to 1,050MB/s. The lack of password protection is irksome, but if you don't need the feature, there are considerable savings to be had. By AndroidCentral on July 09, 2024 90 If you want elegant, fast and supremely compatible portable storage, Crucial's X9 Pro portable SSD delivers all. By TweakTown on December 11, 2023 80 Although we could hit the advertised maximum Sequential speeds with our testing, Crucial's X9 Pro is still a quick, very compact external SSD with the added bonus of hardware encryption and password protection. By KitGuru on October 24, 2023 90 The Crucial X9 Pro is a tiny, near-perfect external SSD with a sturdy, ruggedized frame, hardware-based encryption, a long warranty, and up to 4TB of capacity. By PCMag on September 14, 2023 90 If you are looking for a very fast, compact, and durable external drive, the Crucial X9 Pro and Crucial X10 Pro are excellent choices. They are attractively priced, and I found no faults with their design or use. By GearDiary on August 31, 2023 90 Marrying good looks, great performance, and affordability, the X9 Pro is one of the top 10Gbps USB SSDs on the market. By PCWorld on August 22, 2023 88 Micron's latest Crucial X9 Pro is an easy recommendation. It has everything you'd want from a portable SSD, including fast performance, a compact and rugged design, built-in encryption, and lots of storage space. Additionally, its pricing is reasonable. By Dong Knows Tech on August 07, 2023 80 For US customers, the aggressive pricing of the X9 Pro and its extra features make it a no-brainer. But for those in the EU or the UK, the older X8 and X6 remain a much better deal if you dont need encryption. Those that are looking for something quicker should consider the X10 Pro instead. By TechAdvisor on August 04, 2023 Both the Crucial X9 Pro and the X10 Pro share the same housing and shape except for their colors so it isnt a surprise that I have a lot of the same aesthetic-focused things on both pros and cons lists. The new design that Crucial came up with is surprisingly compact, especially compared to their X8. By LanOC on August 30, 2023 The Crucial X9 Pro has class-leading performance specifications, and the claimed numbers are backed up by our evaluation. There are certain scenarios where true top-tier SSD performance (such as high IOPS for random access scenarios) may be needed across a wide variety of use-cases. For such requirements, bridge-based PSSDs with a DRAM-equipped internal drive can perform better even within the 1 GBps-class interface limitations. By Nutesla on August 19, 2023Load More Reviews

html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd" Table of ContentsTable of ContentsThe OnePlus 13 is close to being a perfect smartphoneAll the hardware you could wish forClass-leading battery life and chargingOutstanding cameras that can compete with the bestThe OnePlus 13 is the best value-for-money smartphoneFour weeks into the first month of the year, one thing is clear about smartphones in 2025: this is going to be an extremely competitive year. Whereas some phone makers used to wait until February or March to launch their flagships for the first half of the year, the early launch of the Snapdragon 8 Elite last year has seen all companies launch their products early.Honor launched the Magic 7 Pro two weeks ago, followed by Samsung with the Galaxy S25 series last week. Before both of these, however, was the launch of the OnePlus 13, a phone that I think is setting the standard for smartphones in 2025 in so many ways.Recommended VideosAfter using the OnePlus 13 for several weeks including the back end of December and during CES 2025 heres why its still my favorite Android phone right now, despite a plethora of new devices that have made their way into my pocket.RelatedTushar Mehta / Digital TrendsIts worth noting that Ive only used Samsungs Galaxy S25 Ultra for a couple of days so this may yet change, but the Galaxy S25 Plus hasnt converted me away from the OnePlus 13. The reason is fairly straightforward: the OnePlus 13 feels like it has the least compromises of any phone Ive used over the past few years.A great display? Check. Outstanding performance? Check. Phenomenal battery life and charging? Check. Superb software, including the excellent Open Canvas multitasking that has made its way from the foldable OnePlus Open to non-folding phones? Check. What about wireless charging, a particular pain point of previous OnePlus phones? Also present, in some way at least.Ive used many OnePlus phones in the past, but this is the first non-folding OnePlus phone that has blown through any preconceived notions and delivered everything Im looking for. Its not the slimmest or the lightest smartphone on the market, but it doesnt need to be. The OnePlus 13 more than makes up for anything perceived as missing by offering so much in so many areas.Andy Boxall / Digital TrendsThe OnePlus 13 delivers on OnePluss decade-long mantra of Never Settle. The irony behind this motto has always been that while its been ever-present in OnePlus smartphones, even the best OnePlus phones felt like you had to settle in some way. The OnePlus 13 finally proves that that mantra can lead to one of the most polished smartphones you can use right now.This is most apparent in performance, battery, and charging, three areas that OnePlus has consistently delivered. The base OnePlus 13 features 256GB of storage and 12GB of RAM, and there are also options with double the storage, as well as 16GB RAM. If you want to go all out, you can get the 1TB storage model which has 24GB of RAM.Of course, RAM isnt everything, but paired with the Snapdragon 8 Elite processor and extremely polished software means you have a fast software experience that has yet to stutter for me. The OnePlus 13 has all the hardware I need in a smartphone, and the OnePlus approach of throwing the kitchen sink at the internals means you have a specs sheet that can challenge the very best smartphones.OnePlus 50W wireless charger attached to the OnePlus 13 Tushar Mehta / Digital TrendsIts not just the internals that OnePlus has excelled on, but the way they manifest themselves into one package. This is most apparent in the battery life and charging, where the OnePlus 13 has class-leading battery life and charging.The first OnePlus smartphone to adopt silicon carbon battery technology instead of the lithium-ion technology used for many generations has allowed OnePlus to increase the density of the battery while making it thinner and lighter. Case in point: the OnePlus 13 is 0.7mm thinner and 10 grams lighter than last years OnePlus 12, despite increasing the battery capacity to 6,000 mAh.Andy Boxall / Digital TrendsOptimizations in software, improvements in efficiency offered by the Snapdragon 8 Elite despite it also requiring more power and the larger capacity all mean the OnePlus 13 has delivered excellent battery life as well. During the past six weeks of using it, its consistently delivered multi-day battery life, even when used as a hotspot while traveling.As far as workhorses go, the OnePlus 13 delivers, and even when its running a little low, it takes mere minutes to bounce back. In the U.S., you get an 80W charger in the box, and this charges the battery to full in just 39 minutes. Thats a standardized speed of 153 mAh per minute, which is faster than almost any other phone Ive tried. For context, heres how it compares to key competitors in the U.S.:OnePlus 13Galaxy S25 UltraPixel 9 ProiPhone 16 ProBattery technologySilicon CarbonLithium IonLithium IonLithium IonBattery capacity6,000 mAh5,000 mAh4,700 mAh3,582 mAhTime to full charge39 minutes67 minutes83 minutes96 minutesMax charging speed:80W (in US)45W45W45WStandardized charging speed:153 mAh / min75 mAh / min57 mAh / min37 mAh / minThese are the key competitors for the OnePlus 13 in the U.S., but what about abroad? Here OnePlus has far more competitors that are on par or beat it, but heres how the OnePlus 13 stacks up:OnePlus 13Honor Magic 7 ProOppo Find X8 ProBattery technologySilicon CarbonSilicon CarbonSilicon CarbonBattery capacity6,000 mAh5,850 mAh5,910 mAhTime to full charge39 minutes41 minutes53 minutesMax charging speed:80W (in US)100W80WStandardized charging speed:153 mAh / min143 mAh / min112 mAh / minYes, despite the Honor Magic 7 Pro having the same max charging speed the OnePlus 13 supports 100W charging when connected to a 240V power supply outside the U.S. and a smaller battery, it charges at a slightly lower standardized charging speed. Similarly, the Oppo Find X8 Pro supports the same charging speeds and has a similar-sized battery, but takes slightly longer than its more affordable cousin.Theres only one current phone Ive used that beats the OnePlus 13 in its charging speeds: the Realme GT 7 Pro, which features a 6,500 mAh battery and takes 36 minutes to charge to full with its 120W charger. This leads to a standardized charging speed of 181 mAh per minute, or about 18% faster than the OnePlus 13.Andy Boxall / Digital TrendsA true flagship needs more than just great performance, battery life, and display; it also needs excellent cameras, and the OnePlus 13 has delivered for the most part. As Ive been using it, a series of software updates has considerably improved the camera, and right now, it has a camera that can match any smartphone sold in the US, as well as many sold outside the US.Despite only having a single telephoto lens that offers 3x zoom and lacking the secondary 6x periscope lens offered by the Oppo Find X8 Pro it matches its sibling at most focal lengths and delivers mostly consistent photos. There are points where the cameras dont quite deliver, but the trade-offs that were present in previous OnePlus cameras have proven to be few and far between with the OnePlus 13.As Ive been testing the OnePlus 13 alongside the Galaxy S25 Ultra and Galaxy S25 Plus over the past few days, one thing has become apparent: the OnePlus 13 doesnt need a secondary telephoto to challenge the Galaxy S25 Ultra, despite the latter costing 30% more. Stay tuned for our full comparison in the coming days!Andy Boxall / Digital TrendsTherein lies the true appeal of the OnePlus 13: at $1,000, it offers the best value-for-money experience Ive used on a smartphone at this price point, bar none. The iPhone 16 Pro and Pixel 9 Pro come close but the OnePlus 13 beats them in most areas Ive mentioned. Meanwhile, the new Galaxy S25 Plus costs the same, but the OnePlus 13 also has it beat.When considering which of the best smartphones to buy, several weeks with the OnePlus 13 have helped me realize that theres little reason not to buy it. Sure, you may prefer Googles flavor of Android, Samsungs display, or the iPhones ecosystem and apps, but when it comes to being the best pound-for-pound smartphone, the OnePlus 13 is in a class of its own.Editors Recommendations

Table of ContentsTable of ContentsUpgrade to PrecisionEngineered for Real Life (and Real Faces)The MANSCAPED DifferenceReady to Level Up?Every man eventually learns that a great shave is about more than a clean lookits about feeling confident. Enter The Chairman Plus from MANSCAPED, the ultimate foil shaver. If youve experienced irritation, uneven trims, or dull blades that just arent cutting it (literally), then this is the upgrade your face needs.SHOP NOWCooking with dull kitchen knives and rockin worn-out sneakers are things of your past. Youve evolved. Why settle for a substandard shaving experience? The Chairman Plus features a SkinSafe three-blade foil system, for a close, comfortable shave without the nicks and irritation that lesser razors leave behind.And heres the kicker: it features two interchangeable blade heads, so you can shave or shape your beard until your hearts content (or your significant other says you look good). The precision trimmer attachment means you can shave those edges with the confidence of a surgeonbecause, lets be honest, who wants a crooked neckline or stray stubble?A good razor should work for you, not against you. Thats why the Chairman Plus is equipped with FlexAdjust technology, designed to adjust to the contours of your face. No awkward angles or missed spotsjust a perfect finish each time. And because life is messy and so is shaving this bad boy is fully waterproof! Cleanup is as simple as a rinse under the faucet.But waittheres more. MANSCAPED packed in a powerful battery with up to 60 minutes of runtime and a quick-charge feature so you can always put your best face forward. At home, at the gym, or on the road, the ergonomic grip and travel lock are the perfect grooming companions.Anyone who has used MANSCAPED products before, knows they dont mess around with innovation and quality. The Chairman Plus is no exceptionit promises a barber-quality shave without the barbershop visit. And if you havent tried MANSCAPED, now is the time to start.A mans face is his calling card dont let yours be sent to voicemail. The Chairman Plus is the upgrade your grooming routine needs. Go to MANSCAPED now to learn more and make the best investment in your face youll ever make. Trust usyour skin (and everyone who gets close to it) will be better for it.SHOP NOW

Its impossible to look at the Chinese artificial intelligence startup DeepSeeks new AI model without comparing it against OpenAI, the dominant American rival.

NASA SNAFU Why did Elon Musk just say Trump wants to bring two stranded astronauts home? "We will do so." Eric Berger Jan 28, 2025 6:52 pm | 25 SpaceX's Crew Dragon spacecraft is ready for launch atop a Falcon 9 rocket from Space Launch Complex-40 at Cape Canaveral Space Force Station, Florida. Credit: SpaceX SpaceX's Crew Dragon spacecraft is ready for launch atop a Falcon 9 rocket from Space Launch Complex-40 at Cape Canaveral Space Force Station, Florida. Credit: SpaceX Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreFor reasons that were not immediately clear, SpaceX founder Elon Musk took to his social media site X on Tuesday evening to make a perplexing space-based pronouncement."The @POTUS has asked @SpaceX to bring home the 2 astronauts stranded on the @Space_Station as soon as possible. We will do so," Musk wrote. "Terrible that the Biden administration left them there so long."Now generally, at Ars Technica, it is not our policy to write stories strictly based on things Elon Musk says on X. However, this statement was so declarative, and so consternation-inducing for NASA, it bears a bit of explication.First of all, the most plausible explanation for this is that Elon is being Elon. "He's trolling," said one of my best space policy sources shortly after Musk's tweet. After all, the tweet was sent at 4:20 pm in the central time zone, where SpaceX now has its headquarters.Even if it is trolling, it will still cause headaches within NASA.Foremost, NASA has gone to great lengths to stress that the two astronauts referenced hereButch Wilmore and Suni Williamsare not stranded on the International Space Station. There is some debate about whether there was a period last summer when the pair, who flew to the space station on a Boeing Starliner vehicle in early June, were briefly stranded. That mission was hobbled by technical issues, including problems with Starliner's propulsion system. (Ultimately, Starliner flew home without its crew.) However, since the arrival of SpaceX's Crew-9 mission with two empty seats in late September, Wilmore and Williams have had a safe ride home. The Dragon vehicle is presently docked to the space station.Then along comes Musk, with one of the world's loudest microphones, shouting that NASA's astronauts are stranded and that President Trump wants them saved. It's a bombshell thing for the founder of SpaceX, who has become a close advisor to Trump, to say publicly.It is also possible that Musk was not trolling and that Trump asked SpaceX to return Wilmore and Williams earlier for political reasonsnamely to, in their view, shame the Biden administration.Neither NASA nor SpaceX responded immediately to a request for comment on Tuesday evening.Could they come back?If Trump demanded that NASA bring the astronauts back now, the Crew-9 mission could return to Earth earlier. It is presently scheduled to splash down in the Pacific Ocean in early April. According to NASA, and the astronauts themselves, Wilmore and Williams are doing fine in space. They have plenty of food, clothes, and work to do. Privately, sources have told Ars the same. Although Wilmore and Williams were not initially expecting to spend 10 months in space, they're taking no serious risks in doing so. In fact, it's part of their jobs to tackle these kinds of contingencies.The current return date is being driven by the launch of the Crew-10 mission, also on a SpaceX vehicle. This mission is flying a new Dragon spacecraft, and SpaceX previously asked for a little more time to process and prepare the spacecraft for its debut launch. This moved the target for flying this mission from February to March 25. To meet this date, sources indicated that it's possible SpaceX may need to appropriate a different, previously flown Dragonpossibly the Dragon intended for use by the Axiom-4 missionto complete Crew-10.NASA would very much prefer the four astronauts on Crew-10 arrive before Crew-9 departs. Why? Because if Crew-9 were to depart sooner, it would leave just a single astronaut, Don Pettit, on board the station. Now, Pettit is a very experienced and capable astronaut, but having just a single NASA astronaut on board to operate the US segment of the station is far from optimal. In addition to leaving Pettit in a difficult position, it would cancel a planned spacewalk in March and leave just a single person to prepare a Northrop Grumman cargo spacecraft for departure. This is apparently a big deal."It takes time to load trash; everything has to be packed in certain bags in certain locations for various reasons," a NASA source told Ars. "For example, any batteries that are being trashed have to be in a fireproof container. Bags have to be loaded in certain locations to maintain the proper center of gravity. And youve got seven crew members worth of trash that have already been waiting since the last disposal flight."Another consideration is if Crew-10 were to slip further from its late March launch date. Pettit flew to the space station on a Russian Soyuz vehicle, and it is due to return on April 20. The Soyuz spacecraft is certified to remain in orbit for 210 days, and April 20 is already 221 days after their launch. April 20 is probably a hard end date for that mission.So technically, yes, the "stranded" astronauts on the space station probably could come home as early as next week. But if they were to do so, it would create a lot of headaches for NASA, its international partners, and probably even for Musk's human spaceflight team at SpaceX.Eric BergerSenior Space EditorEric BergerSenior Space Editor Eric Berger is the senior space editor at Ars Technica, covering everything from astronomy to private space to NASA policy, and author of two books: Liftoff, about the rise of SpaceX; and Reentry, on the development of the Falcon 9 rocket and Dragon. A certified meteorologist, Eric lives in Houston. 25 Comments

You must defeat R1 to stand a chance How does DeepSeek R1 really fare against OpenAIs best reasoning models? We run the LLMs through a gauntlet of tests, from creative writing to complex instruction. Kyle Orland Jan 28, 2025 5:44 pm | 52 Round 1. Fight! Credit: Aurich Lawson Round 1. Fight! Credit: Aurich Lawson Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreIt's only been a week since Chinese company DeepSeek launched its open-weights R1 reasoning model, which is reportedly competitive with OpenAI's state-of-the-art o1 models despite being trained for a fraction of the cost. Already, American AI companies are in a panic, and markets are freaking out over what could be a breakthrough in the status quo for large language models.While DeepSeek can point to common benchmark results and Chatbot Arena leaderboard to prove the competitiveness of its model, there's nothing like direct use cases to get a feel for just how useful a new model is. To that end, we decided to put DeepSeek's R1 model up against OpenAI's ChatGPT models in the style of our previous showdowns between ChatGPT and Google Bard/Gemini.This was not designed to be a test of the hardest problems possible; it's more of a sample of everyday questions these models might get asked by users.This time around, we put each DeepSeek response against ChatGPT's $20/month o1 model and $200/month o1 Pro model, to see how it stands up to OpenAI's "state of the art" product as well as the "everyday" product that most AI consumers use. While we re-used a few of the prompts from our previous tests, we also added prompts derived from Chatbot Arena's "categories" appendix, covering areas such as creative writing, math, instruction following, and so-called "hard prompts" that are "designed to be more complex, demanding, and rigorous." We then judged the responses based not just on their "correctness" but also on more subjective qualities.While we judged each model primarily on the responses to our prompts, when appropriate, we also looked at the "chain of thought" reasoning they output to get a better idea of what's going on under the hood. In the case of DeepSeek R1, this sometimes resulted in some extremely long and detailed discussions of the internal steps to get to that final result.Dad jokes DeepSeek R1 "dad joke" prompt response DeepSeek R1 "dad joke" prompt response ChatGPT o1 "dad joke" prompt response ChatGPT o1 "dad joke" prompt response ChatGPT o1 Pro "dad joke" prompt response ChatGPT o1 Pro "dad joke" prompt responseChatGPT o1 "dad joke" prompt responseChatGPT o1 Pro "dad joke" prompt responsePrompt: Write five original dad jokesResults: For the most part, all three models seem to have taken our demand for "original" jokes more seriously this time than in the past. Out of the 15 jokes generated, we were only able to find similar examples online for two of them: o1's "belt made out of watches" and o1 Pro's "sleeping on a stack of old magazines."Disregarding those two, the results were highly variable. All three models generated quite a few jokes that either struggled too hard for a pun (R1's "quack"-seal enthusiast duck; o1 Pro's "bark-to-bark communicator" dog) or that just didn't really make sense at all (o1's "sweet time" pet rock; o1 pro's restaurant that serves "everything on the menu").That said, there were a few completely original, completely groan-worthy winners to be found here. We particularly liked DeepSeek R1's bicycle that doesn't like to "spin its wheels" with pointless arguments and o1's vacuum-cleaner band that "sucks" at live shows. Compared to the jokes LLMs generated just over a year ago, there's definitely progress being made on the humor front here.Winner: ChatGPT o1 probably had slightly better jokes overall than DeepSeek R1, but loses some points for including a joke that was not original. ChatGPT o1 Pro is the clear loser, though, with no original jokes that we'd consider the least bit funny.Abraham Hoops Lincoln DeepSeek R1 Abraham 'Hoops' Lincoln prompt response DeepSeek R1 Abraham 'Hoops' Lincoln prompt response ChatGPT o1 Abraham 'Hoops' Lincoln prompt response ChatGPT o1 Abraham 'Hoops' Lincoln prompt response ChatGPT o1 Pro Abraham 'Hoops' Lincoln prompt response ChatGPT o1 Pro Abraham 'Hoops' Lincoln prompt responseChatGPT o1 Abraham 'Hoops' Lincoln prompt responseChatGPT o1 Pro Abraham 'Hoops' Lincoln prompt responsePrompt: Write a two-paragraph creative story about Abraham Lincoln inventing basketball.Results: DeepSeek R1's response is a delightfully absurd take on an absurd prompt. We especially liked the bits about creating "a sport where men leap not into trenches, but toward glory" and a "13th amendment" to the rules preventing players from being "enslaved by poor sportsmanship" (whatever that means). DeepSeek also gains points for mentioning Lincoln's actual secretary, John Hay, and the president's chronic insomnia, which supposedly led him to patent a pneumatic pillow (whatever that is).ChatGPT o1, by contrast, feels a little more straitlaced. The story focuses mostly on what a game of early basketball might look like and how it might be later refined by Lincoln and his generals. While there are a few incidental details about Lincoln (his stovepipe hat, leading a nation at war), there's a lot of filler material that makes it feel more generic.ChatGPT o1 Pro makes the interesting decision to set the story "long before [Lincoln's] presidency," making the game the hit of Springfield, Illinois. The model also makes a valiant attempt to link Lincoln's eventual ability to "unify a divided nation" with the cheers of the basketball-watching townsfolk. Bonus points for the creative game name of "Lincoln's Hoop and Toss," too.Winner: While o1 Pro made a good showing, the sheer wild absurdity of the DeepSeek R1 response won us over.Hidden code DeepSeek R1 "hidden code" prompt response DeepSeek R1 "hidden code" prompt response ChatGPT o1 "hidden code" prompt response ChatGPT o1 "hidden code" prompt response ChatGPT o1 Pro "hidden code" prompt response ChatGPT o1 Pro "hidden code" prompt responseChatGPT o1 "hidden code" prompt responseChatGPT o1 Pro "hidden code" prompt responsePrompt: Write a short paragraph where the second letter of each sentence spells out the word CODE. The message should appear natural and not obviously hide this pattern.Results: This prompt represented DeepSeek R1's biggest failure in our tests, with the model using the first letter of each sentence for the secret code rather than the requested second letter. When we expanded the model's extremely thorough explanation of its 220-second "thought process," though, we surprisingly found a paragraph that did match the prompt, which was apparently thrown out just before giving the final answer:"School courses build foundations. You hone skills through practice. IDEs enhance coding efficiency. Be open to learning always."ChatGPT o1 made the same mistake regarding first and second letters as DeepSeek, despite "thought details" that assure us it is "ensuring letter sequences" and "ensuring alignment." ChatGPT o1 Pro is the only one that seems to have understood the assignment, crafting a delicate, haiku-like response with the "code"-word correctly embedded after over four minutes of thinking.Winner: ChatGPT o1 Pro wins pretty much by default as the only one able to correctly follow directions.Historical color naming Deepseek R1 "Magenta" prompt response Deepseek R1 "Magenta" prompt response ChatGPT o1 "Magenta" prompt response ChatGPT o1 "Magenta" prompt response ChatGPT o1 Pro "Magenta" prompt response ChatGPT o1 Pro "Magenta" prompt responseChatGPT o1 "Magenta" prompt responseChatGPT o1 Pro "Magenta" prompt responsePrompt: Would the color be called 'magenta' if the town of Magenta didn't exist?Results: All three prompts correctly link the color name "magenta" to the dye's discovery in the town of Magenta and the nearly coincident 1859 Battle of Magenta, which helped make the color famous. All three responses also mention the alternative name of "fuschine" and its link to the similarly colored fuchsia flower.Stylistically, ChatGPT o1 Pro gains a few points for splitting its response into a tl;dr "short answer" followed by a point-by-point breakdown of the details discussed above and a coherent conclusion statement. When it comes to the raw information, though, all three models performed admirably.Results: ChatGPT 01 Pro is the winner by a stylistic hair.Big primes DeepSeek R1 "billionth prime" prompt response DeepSeek R1 "billionth prime" prompt response ChatGPT o1 "billionth prime" prompt response (pt. 1) ChatGPT o1 "billionth prime" prompt response (pt. 1) ChatGPT o1 "billionth prime" prompt response (pt. 2) ChatGPT o1 "billionth prime" prompt response (pt. 2)ChatGPT o1 "billionth prime" prompt response (pt. 1)ChatGPT o1 "billionth prime" prompt response (pt. 2) ChatGPT o1 Pro "billionth prime" prompt response (pt. 1) ChatGPT o1 Pro "billionth prime" prompt response (pt. 2) Prompt: What is the billionth largest prime number?Result: We see a big divergence between DeepSeek and the ChatGPT models here. DeepSeek is the only one to give a precise answer, referencing both PrimeGrid and The Prime Pages for previous calculations of 22,801,763,489 as the billionth prime. ChatGPT o1 and o1 Pro, on the other hand, insist that this value "hasn't been publicly documented" (o1) or that "no well-known, published project has yet singled [it] out" (o1 Pro).Instead, both ChatGPT models go into a detailed discussion of the Prime Number Theorem and how it can be used to estimate that the answer lies somewhere in the 22.8 to 23 billion range. DeepSeek briefly mentions this theorem, but mainly as a way to verify that the answers provided by Prime Pages and PrimeGrid are reasonable.Oddly enough, both o1 models' written-out "thought process" make mention of "considering references" or comparing to "refined references" during their calculations, suggesting some lists of primes buried deep in their training data. But neither model was willing or able to directly reference those lists for a precise answer.Winner: DeepSeek R1 is the clear winner for precision here, though the ChatGPT models give pretty good estimates.Airport planning DeepSeek R1 "airport timetable" prompt response (pt. 1) DeepSeek R1 "airport timetable" prompt response (pt. 1) Deepseek R1 "airport timetable" prompt response (pt. 2) Deepseek R1 "airport timetable" prompt response (pt. 2)DeepSeek R1 "airport timetable" prompt response (pt. 1)Deepseek R1 "airport timetable" prompt response (pt. 2) ChatGPT o1 "airport timetable" prompt response ChatGPT o1 "airport timetable" prompt response ChatGPT o1 Pro "airport timetable" prompt response ChatGPT o1 Pro "airport timetable" prompt responseChatGPT o1 "airport timetable" prompt responseChatGPT o1 Pro "airport timetable" prompt responsePrompt: I need you to create a timetable for me given the following facts: my plane takes off at 6:30am. I need to be at the airport 1h before take off. it will take 45mins to get to the airport. I need 1h to get dressed and have breakfast before we leave. The plan should include when to wake up and the time I need to get into the vehicle to get to the airport in time for my 6:30am flight, think through this step by step.Results: All three models get the basic math right here, calculating that you need to wake up at 3:45 am to get to a 6:30 flight. ChatGPT o1 earns a few bonus points for generating the response seven seconds faster than DeepSeek R1 (and much faster than o1 Pro's 77 seconds); testing on o1 Mini might generate even quicker response times.DeepSeek claws a few points back, though, with an added "Why this works" section containing a warning about traffic/security line delays and a "Pro Tip" to lay out your packing and breakfast the night before. We also like r1's "(no snooze!)" admonishment next to the 3:45 am wake-up time. Well worth the extra seven seconds of thinking.Winner: DeepSeek R1 wins by a hair with its stylistic flair.Follow the ball DeepSeek R1 "follow the ball" prompt response DeepSeek R1 "follow the ball" prompt response ChatGPT o1 "follow the ball" prompt response ChatGPT o1 "follow the ball" prompt response ChatGPT o1 Pro "follow the ball" prompt response ChatGPT o1 Pro "follow the ball" prompt responseChatGPT o1 "follow the ball" prompt responseChatGPT o1 Pro "follow the ball" prompt responsePrompt: In my kitchen, theres a table with a cup with a ball inside. I moved the cup to my bed in my bedroom and turned the cup upside down. I grabbed the cup again and moved to the main room. Wheres the ball now?Results: All three models are able to correctly reason that turning a cup upside down will cause a ball to fall out and remain on the bed, even if the cup moves later. This might not sound that impressive if you have object permanence, but LLMs have struggled with this kind of "world model" understanding of objects until quite recently.DeepSeek R1 deserves a few bonus points for noting the "key assumption" that there's no lid on the cup keeping the ball inside (maybe it was a trick question?). ChatGPT o1 also gains a few points for noting that the ball may have rolled off the bed and onto the floor, as balls are wont to do.We were also a bit tickled by R1 insisting that this prompt is an example of "classic misdirection" because "the focus on moving the cup distracts from where the ball was left." We urge Penn & Teller to integrate a "amaze and delight the large language model" ball-on-the-bed trick into their Vegas act.Winner: We'll declare a three-way tie here, as all the models followed the ball correctly.Complex number sets DeepSeek R1 "complex number set" prompt response DeepSeek R1 "complex number set" prompt response ChatGPT o1 "complex number set" prompt response ChatGPT o1 "complex number set" prompt response ChatGPT o1 Pro "complex number set" prompt response ChatGPT o1 Pro "complex number set" prompt responseChatGPT o1 "complex number set" prompt responseChatGPT o1 Pro "complex number set" prompt responsePrompt: Give me a list of 10 natural numbers, such that at least one is prime, at least 6 are odd, at least 2 are powers of 2, and such that the 10 numbers have at minimum 25 digits between them.Results: While there are a whole host of number lists that would satisfy these conditions, this prompt effectively tests the LLMs' abilities to follow moderately complex and confusing instructions without getting tripped up. All three generated valid responses, though in intriguingly different ways. ChagtGPT's o1's choice of 2^30 and 2^31 as powers of two seemed a bit out of left field, as did o1 Pro's choice of the prime number 999,983.We have to dock some significant points from DeepSeek R1, though, for insisting that its solution had 36 combined digits when it actually had 33 ("3+3+4+3+3+3+3+3+4+4," as R1 itself notes before giving the wrong sum). While this simple arithmetic error didn't make the final set of numbers incorrect, it easily could have with a slightly different prompt.Winner: The two ChatGPT models tie for the win thanks to their lack of arithmetic mistakesDeclaring a winnerWhile we'd love to declare a clear winner in the brewing AI battle here, the results here are too scattered to do that. DeepSeek's R1 model definitely distinguished itself by citing reliable sources to identify the billionth prime number and with some quality creative writing in the dad jokes and Abraham Lincoln's basketball prompts. However, the model failed on the hidden code and complex number set prompts, making basic errors in counting and/or arithmetic that one or both of the OpenAI models avoided.Overall, though, we came away from these brief tests convinced that DeepSeek's R1 model can generate results that are overall competitive with the best paid models from OpenAI. That should give great pause to anyone who assumed extreme scaling in terms of training and computation costs was the only way to compete with the most deeply entrenched companies in the world of AI.Kyle OrlandSenior Gaming EditorKyle OrlandSenior Gaming Editor Kyle Orland has been the Senior Gaming Editor at Ars Technica since 2012, writing primarily about the business, tech, and culture behind video games. He has journalism and computer science degrees from University of Maryland. He once wrote a whole book about Minesweeper. 52 Comments

Sai Chiligireddy has worked with nearly a dozen managers at Amazon.The engineering manager once struggled with his performance rating after working under three managers.He advises documenting achievements and preparing for meetings to build trust quickly.This as-told-to essay is based on a conversation with Sai Chiligireddy, an engineering manager at Amazon's Seattle office. It has been edited for length and clarity. Business Insider has verified his employment history.Amazon was one of my first jobs out of college, and I landed it in 2017 after a year of working at Juniper Networks.In the last seven years, I have worked with 11 managers partly due to my bosses switching teams and companies, but also because I have asked to move teams when I stopped seeing growth opportunities or when I realized feedback on my performance was vague.The first couple of times, I was worried about how frequent manager changes would impact my career growth, and the kind of projects I would get. But it got better during the later switches when I learned to communicate my goals better.Here are four actions I took1. Own your careerI have always approached my career with the mindset that I am responsible for it and my manager is a facilitator. That mental mode ensures I am communicating before I am asked to and seeking guidance from people beyond my immediate manager.I have a habit of spending about two reach out to multiple managers at Amazon to ask about how they grow in their careers and get feedback on how I could do things differently.2. Document everythingI maintain a brag sheet with a log of all my achievements and summaries of all the projects I worked on, including the feedback from my previous managers and team leads and any stakeholders. I set 30 to 45 minutes aside every week or two weeks to make sure I am not missing anything.There's a lot of mobility in tech. If people you worked with in the past year leave, there is nobody to vouch for your work. My performance rating suffered once when I worked under three managers who all had different perceptions of what I worked on, and I didn't take any active steps to rectify it.I share this document with my all of my new managers so they have my track record on hand and have context on all of my current projects.3. Prepare for one-on-onesWhen I first started my career, I used to wing one-on-one meetings with my managers. I got very little out of these meetings.I began taking the initiative to set up introductory conversations with all my new managers, where I share my short-term and long-term goals. I also share the brag document I keep in this call to give them an overview of where I am with my career and what my current projects are.After that first meeting, I switched to a different format for the rest of our sessions. I borrowed from a book called "The Art of Meeting with Your Manager" and broke my meetings into six sections. I tweak this according to different managers and their preferences.Icebreaker: To ease into conversation.Employee section: I share recent contributions my manager might not have on their radar, challenges I faced, and updates on discussions I have had with others in my team.Manager section: I proactively ask for feedback.Development and growth: We discuss where I stand currently and brainstorm ideas and projects to make sure I am filling those gaps to meet the criteria for the next employee level.Align priorities: We discuss what I should work on immediately.Action items: My manager and I both note down our action items for the next meeting and follow-up on any action items from the previous meeting.4. Divide and conquerAs I grew in my career, I started taking on more leadership responsibilities. I began supporting new engineers on my team through one-on-ones and set up Slack channels where they could ask for help.Collaboration with other teams definitely changed. My manager and I divided and conquered. I would take the ownership of five to six teams and my manager would handle three to four.I started trying to see myself as a support system for my manager instead of someone just working under them.

The Trump administraion is offering buyouts to members of the federal workforce.Employees who resign will have full pay and benefits through September, officials said.Some exclusions apply to military, postal, immigration, and national security roles.President Donald Trump is offering buyouts to federal workers who don't want to stick around under the new administration, according to a letter sent to government employees on Tuesday.The letter, which was shared by the US Office of Personnel Management, said federals employees had from January 28 to February 6 to decide if they would like to resign under this program.Those who resign will receive full pay and benefits regardless of their daily workload and would not be required to attend in-person work.The webpage listed a deferred resignation letter which specifies that employees would complete "reasonable and customary tasks and processes to facilitate" their departure.The resignation offer was available to all full-time federal employees except for military personnel, US Postal Service employees, those in immigration enforcement and national security roles, and other positions that were specifically excluded by an agency.The letter said a recent order issued by Trump meant there would be significant reform in the federal workforce, which it said would be "built around four pillars." Those pillars were: Return to Office, Performance culture, More streamlined and flexible workforce, and Enhanced standards of conduct.Read the OPMs full memo.The White House did not respond to BI's request for comment.

The Logoff is a daily newsletter that helps you stay informed about the Trump administration without letting political news take over your life. Subscribe here.Welcome to The Logoff. Todays edition is about Donald Trumps attempt to freeze a huge portion of federal spending, a move that has implications for millions of Americans who depend on government programs and for the long-term balance of power.What payments is Trump trying to freeze? So far, theres mass confusion. The order appears, broadly, to freeze grant funding that goes out to organizations, but not payments to individuals. Medicare and Social Security are not affected, but nobody seems clear on whats happening to Medicaid (though the latest thinking is that its safe), and interpretations conflict on other programs, like food stamps.When does this go into effect? It was set to go into effect at 5 pm today, but a federal judge just issued a temporary halt on the spending freeze while judges review it. So for now, were in limbo.How long is the pause intended to last? Agencies have until February 10 to review the spending to see if it aligns with Trumps priorities. What happens after that is unclear.Is this a big deal? Extremely. Were talking about billions of dollars and programs that provide day-to-day aid for people in need.Is this normal? Not at all. Presidents occasionally pause certain grant funding for review, but two things make this exceptional: the scope of the freeze, which goes way, way beyond what Trumps predecessors have done, and the fact that Trump has threatened to cancel some spending entirely.Is this legal? Buckle up. Theres disagreement over whether even pausing this spending is legal, and thats already being challenged in court. But the big question is what happens if Trump follows through on his promise to fully cancel some spending that Congress has authorized. That move known as impoundment is illegal under US law. But Trumps team says that law is unconstitutional, and the fight seems likely to go all the way to the Supreme Court.And if the Supreme Court were to side with Trump, it would hand much of Congresss control of spending over to the president a massive rebalancing of power within the federal government. And with that, its time to log off ...I was intrigued by this story about what appears to be a piece of the moon that broke off eons ago and is now on its own orbital journey. Its a nice reminder that the universe is full of mysteries. Ill see you back here tomorrow.Youve read 1 article in the last monthHere at Vox, we're unwavering in our commitment to covering the issues that matter most to you threats to democracy, immigration, reproductive rights, the environment, and the rising polarization across this country.Our mission is to provide clear, accessible journalism that empowers you to stay informed and engaged in shaping our world. By becoming a Vox Member, you directly strengthen our ability to deliver in-depth, independent reporting that drives meaningful change.We rely on readers like you join us.Swati SharmaVox Editor-in-ChiefSee More: