One of the original Overwatch's most best-known features, loot boxes, are coming to Overwatch 2 as part of Season 15 on February 18--though you won't be able to buy them directly as was originally the case.Instead, loot boxes will return as part of Overwatch 2's seasonal free and premium battle passes, as well as rewards for weekly and event challenges. Blizzard is looking to bring more transparency to the system as well, giving players more information in regards to item-drop chances and rarity guarantees. An item of Rare or better quality has a chance to drop from every loot box. After opening five loot boxes, players will be guaranteed an Epic-quality item, and after 20 loot boxes, a guaranteed Legendary. Legendary loot boxes, one of which can be earned via the free battle pass (and two through the premium battle pass), have a 100% chance to reward a Legendary item.Continue Reading at GameSpot

From Software is aiming to give players an authentic version of its RPG experience in quick bursts.

An Enhanced Edition of Neverwinter Nights 2 could be in development, according to recent reports. Together with its predecessor, Neverwinter Nights 2is one of the most well-known role-playing games set in Dungeons and Dragons Forgotten Realms, and this rumor could be excellent news for RPG fans if confirmed.

Stepping deeper into 2025, prospects are looking good for Legend of Zelda fans. Although the Nintendo Switch is passing its torch to the Switch 2, it has had a great streak of Zelda titles, including the industry-shaking The Legend of Zelda: Breath of the Wild near launch, Tears of the Kingdom several years later, and Echoes of Wisdom in what may be its last full year. Even if the Zelda hits stopped at Breath of the Wild, that would be reason enough for celebration, but it went on to include several Nintendo Switch Online ports and other titles like the Links Awakening remake.

Just as there are widely understood empirical laws of nature for example, what goes up must come down, or every action has an equal and opposite reaction the field of AI was long defined by a single idea: that more compute, more training data and more parameters makes a better AI model.However, AI has since grown to need three distinct laws that describe how applying compute resources in different ways impacts model performance. Together, these AI scaling laws pretraining scaling, post-training scaling and test-time scaling, also called long thinking reflect how the field has evolved with techniques to use additional compute in a wide variety of increasingly complex AI use cases.The recent rise of test-time scaling applying more compute at inference time to improve accuracy has enabled AI reasoning models, a new class of large language models (LLMs) that perform multiple inference passes to work through complex problems, while describing the steps required to solve a task. Test-time scaling requires intensive amounts of computational resources to support AI reasoning, which will drive further demand for accelerated computing.What Is Pretraining Scaling?Pretraining scaling is the original law of AI development. It demonstrated that by increasing training dataset size, model parameter count and computational resources, developers could expect predictable improvements in model intelligence and accuracy.Each of these three elements data, model size, compute is interrelated. Per the pretraining scaling law, outlined in this research paper, when larger models are fed with more data, the overall performance of the models improves. To make this feasible, developers must scale up their compute creating the need for powerful accelerated computing resources to run those larger training workloads.This principle of pretraining scaling led to large models that achieved groundbreaking capabilities. It also spurred major innovations in model architecture, including the rise of billion- and trillion-parameter transformer models, mixture of experts models and new distributed training techniques all demanding significant compute.And the relevance of the pretraining scaling law continues as humans continue to produce growing amounts of multimodal data, this trove of text, images, audio, video and sensor information will be used to train powerful future AI models.Pretraining scaling is the foundational principle of AI development, linking the size of models, datasets and compute to AI gains. Mixture of experts, depicted above, is a popular model architecture for AI training.What Is Post-Training Scaling?Pretraining a large foundation model isnt for everyone it takes significant investment, skilled experts and datasets. But once an organization pretrains and releases a model, they lower the barrier to AI adoption by enabling others to use their pretrained model as a foundation to adapt for their own applications.This post-training process drives additional cumulative demand for accelerated computing across enterprises and the broader developer community. Popular open-source models can have hundreds or thousands of derivative models, trained across numerous domains.Developing this ecosystem of derivative models for a variety of use cases could take around 30x more compute than pretraining the original foundation model.Developing this ecosystem of derivative models for a variety of use cases could take around 30x more compute than pretraining the original foundation model.Post-training techniques can further improve a models specificity and relevance for an organizations desired use case. While pretraining is like sending an AI model to school to learn foundational skills, post-training enhances the model with skills applicable to its intended job. An LLM, for example, could be post-trained to tackle a task like sentiment analysis or translation or understand the jargon of a specific domain, like healthcare or law.The post-training scaling law posits that a pretrained models performance can further improve in computational efficiency, accuracy or domain specificity using techniques including fine-tuning, pruning, quantization, distillation, reinforcement learning and synthetic data augmentation.Fine-tuning uses additional training data to tailor an AI model for specific domains and applications. This can be done using an organizations internal datasets, or with pairs of sample model input and outputs.Distillation requires a pair of AI models: a large, complex teacher model and a lightweight student model. In the most common distillation technique, called offline distillation, the student model learns to mimic the outputs of a pretrained teacher model.Reinforcement learning, or RL, is a machine learning technique that uses a reward model to train an agent to make decisions that align with a specific use case. The agent aims to make decisions that maximize cumulative rewards over time as it interacts with an environment for example, a chatbot LLM that is positively reinforced by thumbs up reactions from users. This technique is known as reinforcement learning from human feedback (RLHF). Another, newer technique, reinforcement learning from AI feedback (RLAIF), instead uses feedback from AI models to guide the learning process, streamlining post-training efforts.Best-of-n sampling generates multiple outputs from a language model and selects the one with the highest reward score based on a reward model. Its often used to improve an AIs outputs without modifying model parameters, offering an alternative to fine-tuning with reinforcement learning.Search methods explore a range of potential decision paths before selecting a final output. This post-training technique can iteratively improve the models responses.To support post-training, developers can use synthetic data to augment or complement their fine-tuning dataset. Supplementing real-world datasets with AI-generated data can help models improve their ability to handle edge cases that are underrepresented or missing in the original training data.Post-training scaling refines pretrained models using techniques like fine-tuning, pruning and distillation to enhance efficiency and task relevance.What Is Test-Time Scaling?LLMs generate quick responses to input prompts. While this process is well suited for getting the right answers to simple questions, it may not work as well when a user poses complex queries. Answering complex questions an essential capability for agentic AI workloads requires the LLM to reason through the question before coming up with an answer.Its similar to the way most humans think when asked to add two plus two, they provide an instant answer, without needing to talk through the fundamentals of addition or integers. But if asked on the spot to develop a business plan that could grow a companys profits by 10%, a person will likely reason through various options and provide a multistep answer.Test-time scaling, also known as long thinking, takes place during inference. Instead of traditional AI models that rapidly generate a one-shot answer to a user prompt, models using this technique allocate extra computational effort during inference, allowing them to reason through multiple potential responses before arriving at the best answer.On tasks like generating complex, customized code for developers, this AI reasoning process can take multiple minutes, or even hours and can easily require over 100x compute for challenging queries compared to a single inference pass on a traditional LLM, which would be highly unlikely to produce a correct answer in response to a complex problem on the first try.This AI reasoning process can take multiple minutes, or even hours and can easily require over 100x compute for challenging queries compared to a single inference pass on a traditional LLM.This test-time compute capability enables AI models to explore different solutions to a problem and break down complex requests into multiple steps in many cases, showing their work to the user as they reason. Studies have found that test-time scaling results in higher-quality responses when AI models are given open-ended prompts that require several reasoning and planning steps.The test-time compute methodology has many approaches, including:Chain-of-thought prompting: Breaking down complex problems into a series of simpler steps.Sampling with majority voting: Generating multiple responses to the same prompt, then selecting the most frequently recurring answer as the final output.Search: Exploring and evaluating multiple paths present in a tree-like structure of responses.Post-training methods like best-of-n sampling can also be used for long thinking during inference to optimize responses in alignment with human preferences or other objectives.Test-time scaling enhances inference by allocating extra compute to improve AI reasoning, enabling models to tackle complex, multi-step problems effectively.How Test-Time Scaling Enables AI ReasoningThe rise of test-time compute unlocks the ability for AI to offer well-reasoned, helpful and more accurate responses to complex, open-ended user queries. These capabilities will be critical for the detailed, multistep reasoning tasks expected of autonomous agentic AI and physical AI applications. Across industries, they could boost efficiency and productivity by providing users with highly capable assistants to accelerate their work.In healthcare, models could use test-time scaling to analyze vast amounts of data and infer how a disease will progress, as well as predict potential complications that could stem from new treatments based on the chemical structure of a drug molecule. Or, it could comb through a database of clinical trials to suggest options that match an individuals disease profile, sharing its reasoning process about the pros and cons of different studies.In retail and supply chain logistics, long thinking can help with the complex decision-making required to address near-term operational challenges and long-term strategic goals. Reasoning techniques can help businesses reduce risk and address scalability challenges by predicting and evaluating multiple scenarios simultaneously which could enable more accurate demand forecasting, streamlined supply chain travel routes, and sourcing decisions that align with an organizations sustainability initiatives.And for global enterprises, this technique could be applied to draft detailed business plans, generate complex code to debug software, or optimize travel routes for delivery trucks, warehouse robots and robotaxis.AI reasoning models are rapidly evolving. OpenAI o1-mini and o3-mini, DeepSeek R1, and Google DeepMinds Gemini 2.0 Flash Thinking were all introduced in the last few weeks, and additional new models are expected to follow soon.Models like these require considerably more compute to reason during inference and generate correct answers to complex questions which means that enterprises need to scale their accelerated computing resources to deliver the next generation of AI reasoning tools that can support complex problem-solving, coding and multistep planning.Learn about the benefits of NVIDIA AI for accelerated inference.

Theres a clip making the rounds today of former Activision Blizzard CEO and all-around video game fandom boogeyman Bobby Kotick recalling a failed acquisition and fumbling the details. On a video podcast for venture capital firm Kleiner Perkins, featuring former EA exec Bing Gordon, Kotick confessed: We actually had a bad acquisition. The company that was, um in Manchester, that did the driving game for Xbox, and it was called, um You can watch the whole thing here, around the 8-minute mark. So yeah, not a great look. But when Im [checks Wikipedia] 62 years old, I suspect I may not remember something I did when I was [checks calculator app] 44 years old, which is basically how old I am now and Im feeling a way about it. Will I remember writing this post? Who knows.But what I do remember is Bizarre Creations absolute showstopper of a game one that my family still plays to this day.No, not Blur all respect to the Blur heads out there and no, not the driving game for Xbox, though respect to the Project Gotham Racing fans as well. Im of course talking about Geometry Wars, the game that helped establish a new kind of console game on Microsofts muscular sophomore effort and, in so doing, became arguably the Xbox 360s best launch title.Originally included inside Project Gotham Racing 2 on the original Xbox, Geometry Wars is a simple twin-stick shooter: left stick controls the ship, and right stick fires. Thats it! In one of my first jobs as a video game journalist (RIP Joystiq), I covered the Xbox 360 launch in the days leading up to its November 2005 release, and Geometry Wars: Retro Evolved as the high-definition Xbox Live Arcade version was called immediately impressed. I wrote, If [Project Gotham Racing 3] isnt the highlight of the 360 launch, Bizzarres other creation surely is!I stand by it!In the weeks and months that followed, the rest of the Joystiq staff and I watched the global leaderboards with some combination of wonder and shock. They scored what now? The novelty of this kind of retro game being played on a high-def game console with the internet baked into its very fabric was the smallest of hints as to what was to come for video games, but it was a potent dose. We reported out high scores (congrats Striker and xTHeDoGgx) and challenged each other on the friends leaderboards.And now, Geometry Wars: Retro Evolved remains a family favorite on the Steam Deck, where its bite-sized sessions coupled with some healthy father-son competition have reminded me how durable this games formula is, almost twenty years later. So even if you dont remember the name of the studio, or the city it was based in Liverpool, by the way you should remember Geometry Wars. Activision did close Bizarre Creations in 2011, following the launch of racing title Blur, and after some various moves, a handful of that team has reconstituted itself at Lucid Games, who managed to release Geometry Wars 3: Dimensions in 2014, marking the franchises first PlayStation release.And if you havent played it, I have good news for you! You can still purchase the original on the Xbox store for the same $4.99 it cost in 2005, or on Steam where its currently going for $1.99 thats less than a gas station coffee! If youre more of a PlayStation person, you can snag Geometry Wars 3 for $14.99 which is basically the price of a really fancy coffee nowadays (dont fact-check me).Im curious if any of you share my fondness for this game; let me know below. See you on the leaderboards.

As Elden Ring Nightreign previews begin to slowly trickle out ahead of its May 30 release date, battle designer-turned-director Junya Ishizaki sat down with GameSpot to discuss some of the more interesting choices the FromSoftware team made with the spin-off, including the decision to incorporate old bosses from the Dark Souls franchise.The primary reason for these existing bosses in Nightreign is from a gameplay perspective, Ishizaki said. Of course, with this new structure and with this new style of game, we needed a lot of different bosses. We needed a lot to add to this mix, so we wanted to leverage what we deemed appropriate from our previous titles.This makes a lot of sense, considering FromSoftwares established reputation for smartly using old assets in new games. They needed bosses for Nightreign, they had a bunch already developed from previous projects, so why not provide a little fan service? Speaking personally, however, Ishizaki offered an even more succinct explanation: I thought itd be kind of fun.Elden Ring Nightreigns reveal ignited a firestorm of discussion. Not only is a new FromSoft game always exciting, but the pre-release teaser also included footage of reused Dark Souls bosses like the Centipede Demon, the Dukes Dear Freja, and the Nameless King.Fans immediately began theorycrafting lore reasons for these throwbacks. Was this evidence, as IGN proposed, of a FromSoft multiverse? Weve seen hints in the past, such as the character Patches showing his sneering face in the studios games for decades, and Nightreign was considered by some to be further proof that these connections were more than a matter of thematic consistency or even just mere Easter eggs.But nope! Sounds more like FromSoft simply needed to fill out the boss roster, and Ishizaki thought bringing in Dark Souls bosses would be fun. As someone experiencing severe multiverse fatigue thanks to an entertainment industry loathe to develop new ideas, Im totally okay with this explanation.In his preview for Polygon, entertainment editor Austen Goslin described Elden Ring Nightreign as a game for the speed demons and the degenerates that feels like a carefully thought out, official version of the randomizer mods that FromSoftware players have used for years to keep the games they love fresh.

We may earn a commission from links on this page.The first time I put on real lifting shoes, it was a revelation. My feet were so stable during squats, it was like they were glued to the ground. They were so comfortable I forgot to take them off for deadlifts, howevera rookie mistake, as the heels put you at a disadvantage for that lift. Heres what you should know about the shoes you wear when you lift heavy.Why your shoes matterWhen you first head into the gym, youll wear whatever shoes youve got. If that means a pair of running shoes, its not the end of the world. But as soon as you can, you should upgrade. Running shoes, and other comfortable sneaker-like footwear, tend to have squishy soles. They're great for comfort and for cushioning your feet during a run. But when you're lifting heavy, that cushion introduces some problems. You won't have as solid a connection to the floor, and if your weight shifts a little, you could find yourself wobbling a bit. This extra instability is not dangerous (that's a myth) but if you lift in running shoes, you're missing an opportunity to give yourself a properly stable base. Do yourself a favor and give yourself the proper footwear for the job. Besides "not squishy," there is one other consideration: How high do you want the heels on your lifting shoes? You have two options hereeither flat-soled shoes like Chucks or Vans, or Olympic weightlifting shoes with a hard, elevated heel (typically 3/4"). It is not weird to bring two pairs of shoes to the gym. Let's look at your options. Heeled weightlifting shoes are for squats (and Olympic lifts)Let's take a look at heeled weightlifting shoes, also called "squat shoes" or "lifters." These aren't commonly sold in sporting goods stores, so if you haven't seen or heard of these, you're not alonebut they are quite common on the feet of serious lifters, and can be ordered online. (I'll recommend a few brands below.) These weightlifting shoes have a non-cushioned sole and a hard, elevated heel, made of plastic or (traditionally) wood. The heel is usually three-quarters of an inch tall, which may not feel like much, but really makes a difference at the bottom of a squat. If you have trouble squatting to your desired depth, you could probably use a little help with ankle mobility. Besides doing ankle mobility exercises, you can also wear heeled shoes for an instant improvement. Even for people with excellent ankle mobility, heeled lifters can be beneficial. Olympic weightlifters, even the ones with freakishly good ankle mobility, all wear weightlifting shoes in competition and for most of their training. The heel helps you to keep a more upright body position at the bottom of the squat, which helps you to be stronger and more stable. (This Instagram post from Catalyst explains more of the reasoning while showing off what kind of shoes I mean.)I compete in Olympic weightlifting myself, and I'll wear my lifters for my entire workout, unless there's something specific that requires a different type of shoe. (Plyometrics in the warmup? Running shoes are best for that.) If your training is more bodybuilding or powerlifting style, you may want to wear flat shoes for most of your lifts, and change into lifters for squats. One bonus for the powerlifters: If you compete in a federation that requires you to keep your whole foot on the floor while bench pressing, the heel on lifting shoes can help. (If you dont compete, or if keeping your heel on the ground isnt an issue for you, then you can bench in whatever shoes you want.) The best weightlifting shoes to considerMost weightlifting shoes perform equally well, and only differ in their fit. You don't need to overthink the question of which shoe is "best" according to other people; just try on a few and choose the one that fits your feet. The shoes should fit snugly, without pinching when you walk around or when you get into a lunge (or split jerk) position. If they hurt, they're too tight. If they flop around or feel like a roomy running shoe, they're too loose.If your feet are narrow, try Nike Romaleos. If your feet are wide, try Rogue Do-Wins. If you prefer a wide toe box (wider in the front of the shoe, slightly narrower in the back), try the TYR L-1. Other popular lifters include Adidas Adipowers and the Reebok Legacy Lifter. Reebok Legacy Lifter $199.95 at Amazon $220.00 Save $20.05 Shop Now Shop Now $199.95 at Amazon $220.00 Save $20.05 Most weightlifting shoes are sold in men's or unisex sizing, so don't worry if you can't find a "women's" version. Typically you'll subtract 1.5 sizes from your women's size to find your men's size, but check the company's size charts to confirm. For an example of sizing, I wear a women's 8 in most shoes, and size up to 8.5 in Nike running shoes. When I bought my Romaleos, I went with the men's 6.5, which corresponds to a women's 8, same as my street shoes. They fit like a glove. Unfortunately, weightlifting shoes tend to be pretty expensiveyou'll pay well over $100 for most models, and $200 is a more typical price. On the bright side, they'll last years if you take care of them and only wear them indoors. (They're not disposable like running shoes, thank God.) There are also plenty of older models and lightly used pairs available on places like eBay. If you're buying new, Do-Wins tend to be the cheapest of the major brands. Flat shoes are for deadlifts and general strength trainingIf you don't need heeled lifters, or if you already have a pair but want something else for other lifts, your best bet is to find a comfortable shoe with a thin, flat sole. Flat shoes are better than running shoes in the gym because they don't have that squishy layer that can make you feel unstable. They aren't quite as stable as the heeled lifters we discussed above, but they're cheaper and more versatile. They are also lower to the ground, which can be an advantage when you're deadlifting. In a deadlift, you're picking a bar up from the floor, and standing up. The taller your heels, the farther you have to move the barand an extra three-quarters of an inch can be enough to make the difference between a successful PR attempt and one that was almost there. (As with running shoes, this isn't a huge problem, but some of us like to take every advantage we can get.) I mentioned above that I wear lifters for my Olympic weightlifting workouts. But if I'm training in any other style, especially if it involves a lot of deadlifts or a mix of movement types (farmer's carries, box jumps) I'll wear my Chucks.The best flat-soled shoes for strength trainingThese are a bit easier to shop for than weightlifting shoes, since a lot of casual shoes fit the bill, and you don't have to stick to athletic brands. The classic choices in this category are Converse (Chuck Taylor high-tops or low-tops) or Vans Old Skools. Chuck Taylor All Star Canvas High Top $63.20 at Amazon $68.95 Save $5.75 Shop Now Shop Now $63.20 at Amazon $68.95 Save $5.75 You can always deadlift in socksI need to mention one simple fix for the "okay, how many shoes do I have to bring to the gym??" problem. You can do your deadliftsand other lifts, if neededin socks. Now, dont deadlift barefoot if its a gym you share with other people. Thats usually against the rules. But socks are fine. You wont have any cushioning under your feet, and you won't have that 3/4" heel either. In competition, lifters sometimes wear thin-soled slippers for deadlifts just so they can technically be wearing shoes but still lift as if they were in socks. Socks are great if you pull conventional (with your feet together), but if you do sumo deadlifts instead, you may find socks too slippery on your lifting platform. If thats the case, get some grippy deadlift slippers, or go back to your trusty Chucks.

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.Ever since my home was broken into, I've installed as many security cameras, floodlights, and other security measures in place as I could get away with (and they've been working great so far). The video doorbell I chose to go with for my front door is the Reolink Video Doorbell. Right now, you can get one of the best video doorbells of 2025 for a $51.99 (originally $109.99), the lowest price it has ever been, even cheaper than used ones on Amazon, according to price-tracking tools. Reolink Video Doorbell Connection: Wifi and POE, Resolution: 2K, Visibility: 180 degrees, Night Vision: Yes. $51.99 at Woot $109.99 Save $58.00 Get Deal Get Deal $51.99 at Woot $109.99 Save $58.00 One of the main reasons I decided to go with the Reolink Video Doorbell is because there is no monthly subscription fee unless you want to pay for cloud servicebut you don't need to, as you can just use an SD card for local storage (it auto-deletes the old footage so the local storage doesn't get full). I loved that it comes with a free plug-in chime so you can hear it wherever you put it in your home. It's also very flexible; you can choose to install it with an ethernet cable if your wifi is weak or wired if you don't want to depend on batteries.The advantage of using an ethernet cable (other than a solid, reliable internet connection) is that it won't be affected by wifi jammers that robbers might use before a break-in. It also has a pre-roll video feature that records six seconds before an event is triggered so you can have context on the event. The field of view is designed to capture a wide 180-degree view, meaning you won't be able to see packages on your porch (unless you angle it down, but you might miss catching people's faces who get close to the door). You can read more about this camera in PCMag's review.

Iconic game designer Jeff Minter is back with another modern take on a long-forgotten Atari title. Minter has turned his psychedelic eye toward the 1984 arcade cabinet I, Robot. His version ups the visuals and takes serious liberties with the original design, adopting techno music and some new game modes.The original I, Robot was a flop, despite being made by Dave Theurer, the guy behind Missile Command and Tempest. Maybe it was just ahead of its time. After all, it was the very first commercial video game to use real-time, flat-shaded 3D polygon graphics. The game came out a full eight years before Segas Virtua Racing and more than a decade before the PlayStation and N64 brought 3D gaming into the mainstream.The original title had players control a robot as it jumped around mazes to flip the color of tiles and shoot enemies. There was also a giant eyeball that had to be avoided at all costs. Minters version keeps the core gameplay mechanic, but increases the speed and incorporates new gameplay elements. For instance, there are new tube shooter levels and an exploration mode called Ungame. It looks pretty nifty.If the name Jeff Minter seems vaguely familiar, hes the person who made Tempest 2000 and its various sequels. More recently, he remade an unreleased Atari prototype called Akka Arrh that originally dates back to 1982. He also provided the visuals for a Nine Inch Nails video.Minters version of I, Robot will be released this spring on pretty much every platform, including Nintendo Switch, PC, PS4, PS5, Xbox One and Xbox Series X/S. The design team is also working on a port for PlayStation VR2.This article originally appeared on Engadget at https://www.engadget.com/gaming/acclaimed-designer-jeff-minter-is-back-with-a-remake-of-the-80s-arcade-curio-i-robot-194037105.html?src=rss