It's no secret that the Marvel Cinematic Universe is in a bit of a rough spot. After the high that was Avengers: Endgame, many thought Marvel Studios could do no wrong, especially as they followed it up with hits like WandaVision, Loki, and Spider-Man: No Way Home. Sadly, a mixture of behind-the-scenes turmoil, box office bombs, and unnecessary Disney+ projects have driven away many fans.

While the Dragon Isles offers up brand-new factions, landscapes, and villains to defeat, for Hunters, there are a variety of interesting new pets just waiting to be tamed. As can be expected, the pets found in World of Warcraft: Dragonflight appear primitive and ancient, just like the continent adventurers find themselves on.

Has this bug been fixed in the post 2.37.0 work so far?https://github.com/conan-io/conan-center-index/issues/25353Although it was reported to that Conan repository, I suspect it has not been reported upstream.I've also run into it with Homebrew Clang 18 and 19.Thanks,Rob

Garmin produces more variations on the smartwatch theme than anyone else, and the Instinct series is the range that prioritizes battery life first and foremost (with a little help from solar charging). Three years after the Instinct 2 series first appeared, Garmin has now unveiled the Instinct 3 models at CES 2025.The headline feature here is an upgraded screen option. One of the main reasons the Instinct 2 watches offer such good battery life is they use low-res, monochrome MIP displays: These displays are still available with the Instinct 3 series, but there are now models with bright, full color AMOLED displays, too (check our guide to MIP vs AMOLED for more details). The new AMOLED option for the Instinct 3. Credit: Michelle Ehrhardt If you go for a monochrome MIP display Instinct 3, the screen is slightly bigger than it was on the Instinct 2, and you also get solar chargingwhich in theory gives you unlimited battery life (assuming you live in a sunny part of the world). Garmin says the solar charging panel has been improved on the new watches, giving you (for example) more than five times the battery life in GPS mode when solar charging on the largest size.Opt for the AMOLED screen, and you don't get solar charging or quite as much battery life, but you do get full color displays. Garmin says the AMOLED models in "smartwatch mode" (with most features switched on but no GPS) can deliver up to 24 days of battery life. For the MIP Solar model, that goes up to 40 daysand if you're always in sunny weather and never pushing your watch too hard, you may never need to recharge it again.As usual, battery life will depend on a variety of factors, not just the screenso take these as estimates rather than guarantees. Like the Instinct 2 ranges, the Instinct 3 smartwatches come with a variety of battery modes so you can decide whether to prioritize features and performance or battery life. The monochrome MIP option for the Instinct 3, matching the Instinct 2. Credit: Michelle Ehrhardt Both monochrome MIP and AMOLED options are available in 45mm or 50mm sizes, and all models have a built-in flashlight and a reinforced metal bezel. First introduced with the Instinct 2X, the flashlight is now available across all the Instinct 3 models, and comes with a clever trick: You can set it to a strobe mode, which might come in handy when you need to make yourself visible while jogging late at night, for instance.There's also a more affordable Instinct E model, available in 40mm and 45mm sizes, which has the monochrome MIP display but no solar charging, no flashlight, and no metal bezel. It does still offer a lot of the same fitness tracking and navigation tricks though, so is worth weighing up if you have a smaller budget and like the Garmin Instinct aesthetic.Of course, all of the usual Instinct benefits are passed over to the new range, including 10 ATM water resistance, and MIL-STD 810 military standard thermal and shock protection. The rather unusual screen layout is retained on the MIP models too (and available as an option on the AMOLED watch faces), with a little screen-within-a-screen in one corner, that can be set up to show a variety of stats or indicators. The new and more affordable Instinct E model. Credit: Michelle Ehrhardt There are no new sensors on the Instinct 3, but there wasn't much need for an upgrade here: As before, the watch can track a host of health and fitness stats (including heart rate variability and blood oxygen saturation), as well as offering GPS and compass navigation, contactless payments, and smart notifications from your phone.Tempted to buy one? Orders open Jan. 10 on the Garmin website, with prices starting at $449.99 for the 45mm AMOLED model, $399.99 for the 45mm MIP Solar model, and $299.99 for the 40mm Instinct E model. The primary color option is Black/Charcoal, but a number of limited edition choices are also going to be available.

The first thing people tend to do when they get a pizza oven is invite you over for a pizza party, at which they will inevitably serve burned, yet undercooked pizza, and maybe even sacrifice a pie or two by dropping it in a failed attempt to turn it while in the oven. At CES 2025, Current Backyard has introduced a smart pizza oven that might solve for these problems. The Current Model P Smart Pizza Oven is an electric pizza oven enabled with wifi and Bluetooth that can be used both indoors and outdoors. With help from an accompanying app, it promises to cook a perfect pie every time.Cook a pizza at up to 850FOn its face, the concept is appealing: An oven that eliminates the most common pains of pizza-making by leveraging smart tech. The Model P uses an algorithm to cycle its heating, and so you never need to turn the pizza while it is in the oven, which is hard to do without a lot of practice. You dont even need to open the door while the pizza is cooking.The accompanying app has a Pizza Build Calculator, so regardless of what kind of pizza you want to make, it will tell you the exact amount of sauce, cheese, and toppings you need for, say, a 12-inch pizza in the style you choose (Neapolitan, New York, Thin Crust). Then, it will tell the oven how long and at what temperature to cook itup to 850F.You just have to get the pizza onto the cordierite cooking stone and then let the app and the oven handle it from there. Current claims that the oven doesn't need to reheat between pizzas, eitheryou can slide them in and out continuously, which is more like how a traditional wood pizza oven operates. Since a pizza oven is still just an oven, you can utilize the Model P for other cooking instances using the low or high broil function.How much the Current smart pizza oven will costThe oven looks sturdy enough, and is made of stainless steel. It comes with a 10-year limited warranty, and at $599, is priced on par for the market when compared to other electric ovens with the same temperature capacity. (While pricing for dedicated pizza ovens varies, theyre often not cheapthis smart model from Breville, for example, starts at around $800).As someone with a few smart cooking devices, including the Brisk It Origin grill and the Brava Oven, both of which work the same way (let the app and device take over the cook time and temperature), I am a fan of the general concepts. I am a confident cook, but it's also nice to sometimes ease up and let a machine take over. The oven isnt making the sauce or choosing the quality of toppingsyoure still ultimately in charge of the quality of the resultbut the labor of standing over a stove and figuring out when something is done cooking is off the table, so to speak.Current has a mixed track record (for me)That said, I tested a Current smart backyard grill this past summer, and although it did precisely what it promised, I didnt love the experience. Electric grills are great if you cant use gas, charcoal, or wood for some reason, but you dont get the same great smokiness or char you would from a real grill. I also found the Current to be a pain to clean. Without hands-on testing, it's hard to know if this pizza oven will offer similar setbacks.As pizza ovens go, 850F isnt actually all that hotwood-fired pizza ovens can climb to over 1000F, and high heat is critical for creating the best crust. Still, it's not like you can bring a wood pizza oven inside your home, and electric pizza ovens are already quite popular, so adding smart tech to one of them is a potential win. We'll find out when it hits the market: The Model P should be available sometime in the second quarter of 2025.

I've seen plenty of weird and wonderful devices at the annual CES expo, alongside more serious and heavyweight hardware launches, but the Nkojita FuFu might be my favorite quirky gadget of the 2025 show yet: The tiny cat-like robot sits on the side of your mugs and blows on your drinks and food to cool them down. It's functional, and also rather adorable.According to Yukai Engineering, the Japanese manufacturer behind the robot, it's "designed to recreate the joy of encountering and interacting with a human baby"presumably because it's small and playful, not because it blows on your beverages. You can put it on the side of any drink or dish of food with a raised edge that's thin enough for the little bot to perch on top of. Credit: Jake Peterson In Japanese, "neko-jita" means "cat tongue," a description used for an intolerance of hot food and drink (a 2018 poll quoted by Yukai Engineering suggests it's something that bothers nearly half the population in Japan). The "fufu" part of the name is intended as an onomatopoeic expression imitating steady breathing.There are actually more smarts here than you might think, in addition to the stylish and cutesy design. The internal fan inside the Nkojita FuFu robot uses a special algorithm that will support up to eight different blowing modes, each with their own custom name: From a gradual ramping up in strength ("Look at That!") to a long, non-stop blow ("'Til You Drop").Apparently not all of these modes are guaranteed to make it to the final product, and it seems the bot will switch between them at randomthe idea being that the cooling gadget will then feel a little less robotic or like a standard fan. It works via an internal battery, which you can recharge via a USB-C port.The development team says their creation will cool hot water down from 190F (88C) to 160F (71C) in three minutes and 151F (66C) in five minutes. Without any additional cooling, the comparison temperatures are 176F (80C) after three minutes and 171F (77C) after five minutes, so you can see how much it speeds up the process. The bot can just sit to the side of a dish. Credit: Yukai Engineering "We created Nkojita FuFu as a little personal meal partner, so you can solve your nko-jita problem anywhere, anytime," says Tsubasa Tominaga, the chief marketing officer at Yukai Engineering. "The robot doesnt just help you enjoy hot coffee. It can make it easier for seniors to down hot soup and reduce parents chores by helping kids eat with less help."The bad news is, the "mascot robot" is unlikely to be available outside of Japan, so if you live anywhere else you might have to get creative when it comes to ordering one. A crowdfunding campaign is scheduled for the middle of 2025, and the final retail price is expected to be 3,800 Japanese yen (roughly $25).

An AU$3 billion equity investment from the Federal Government puts the NBN on track to finish its upgrade plan.

xAI releases Grok mobile app for iOS.

You walk into CES and youll probably see more chips, gadgets, microprocessors, and tech youve ever seen in your life. Amidst this forest of plastic, metal, and silicone, I managed to spot a bit of greenery and obviously I stopped dead in my tracks. You see, Im a bit of a gardener myself, having grown tomatoes, lemons, basil, mint, sunflowers, bell peppers, and now even peas and chillis. I recently invested in a hydroponic tower, that grows plants without soil, instead using only water. The obvious benefit of a hydroponic tower is that its compact, efficient, fast, and resists plant infections because bugs only thrive in soil.On the floor at CES I saw the Gardyn, a hybridponic tower that also adds AI to the mix, giving you a plant tower thats smart enough to grow itself rather than involving you with the nitty-gritty details. Powered by two cameras that point directly at the plants while they grow, the Gardyn automatically adjusts lighting and water distribution to its plants, leveraging the power of artificial intelligence to essentially hack nature into growing at its fastest pace. What youre then left with is a kitchen garden that consistently provides you with fresh produce, while you essentially do the bare minimum.Designer: GardynThe Gardyn is almost like your standard hydroponic tower. It comes with a water reservoir at the bottom, and a vertical channel that houses the plants. Water travels from the bottom to the top, raining down on the pods which contain the seeds. Eventually, once the plant grows, the roots get watered occasionally, and nutrients in the water help the plant grow without any soil. The Gardyn comes with sunlight-mimicking light towers too, which switch on or off automatically in line with the circadian rhythm. A camera on the light rod monitors the plants 24*7, getting a sense of how the plants are doing while informing the AI of the plants needs and their progress.Kelby, the Gardyns AI assistant, watches over your plants with more dedication than most babysitters. Through a built-in camera and sensors, Kelby knows exactly when to water, how to distribute nutrients, and even which plants are thriving or struggling. My favorite detail? It compiles a time-lapse of your plants growing, which feels like a small nod to the magic of nature even as technology orchestrates the whole thing.Heres where things truly get interesting. The Gardyns hybridponic tower uses machine learning to develop faster ways to grow better plants, but it doesnt do so in an isolated manner. It works alongside thousands of other Gardyn devices, operating like a hive mind where each machine shares information to help all the Gardyn devices perform better. If a certain watering schedule benefits Kale more, the camera will record the evidence and that Gardyn will teach other Gardyns to water kale in the most hyper-optimized way.What you essentially have to do is just add these readymade pods into your Gardyn. Gardyn provides pre-seeded pods called yCubes, which eliminate the fiddly, messy step of sprouting seedlings elsewhere. If youve ever wrestled with rock wool and seed trays, this feels like skipping to dessert without eating your vegetables. Just plug the pods into the Gardyn, fill the water tank, and let the system do its thing. A unique code allows the Gardyn to know which seeds are inside the pod, so the tower essentially has a record of what plants youre growing. The machine circulates water at just the right intervals, using 95% less than traditional gardening. Yes, you still have to refill the tank now and then, but thats about as demanding as it gets.The yCubes come in multiple packs, and you can simply install them into the Gardyn when you want to grow a new plant. Some plants (like spinach for example) have a finite life, which means they either die or become unsuitable for consumption. When your plant reaches that stage, all you do is pull the old yCube out and add a new one. The Gardyn identifies that a new pod has been added to the tower and optimizes light and watering to the tower accordingly.The beauty of this device lies in two things firstly, seeing an AI basically turn into an expert horticulturist and farmer, growing your own kitchen garden for you, and secondly, having access to an app that doesnt just give you all the info you need, but also lets you view a time-lapse of your garden, seeing plants grow from saplings to proper flora in a 10-second clip thats perfect for the gram.Gardyns team also really emphasized how having and AI basically grow your produce for you really helps you understand how good vegetables can taste. Apparently, the apple you eat today was probably harvested months ago and stored in a deep freezer so it could be transported across the globe. Unless you buy food from a farmers market, the food youre buying is easily weeks or months old, and are made to artificially ripen using certain chemicals. When you grow your food AT HOME, youre getting the freshest produce ever, and that can have a remarkable effect not just on cost, but also on taste. Lettuce tastes better, tomatoes are sweeter, jalapenos tangier, etc.And all this can basically exist in a tiny area, giving you an entire farm in the floor-space of essentially a doormat. The Gardyn is compact and vertical, taking up just two square feet of floor space, but it can grow up to 30 plantsplenty for herbs, greens, or even bigger crops like tomatoes and eggplants. For smaller households or limited spaces, theres a version with fewer slots. Either way, youre growing a jungle in a corner. A jungle thats self-growing, self-sustaining, and human-nourishing!The post This AI-powered Planter Automatically Grows Up To 30 Plants Without Any Soil: Hands-on at CES 2025 first appeared on Yanko Design.

The rapid growth of digital platforms has brought image safety into sharp focus. Harmful imageryranging from explicit content to depictions of violenceposes significant challenges for content moderation. The proliferation of AI-generated content (AIGC) has exacerbated these challenges, as advanced image-generation models can easily create unsafe visuals. Current safety systems rely heavily on human-labeled datasets, which are both expensive and difficult to scale. Moreover, these systems often struggle to adapt to evolving and complex safety guidelines. An effective solution must address these limitations while ensuring efficient and reliable image safety assessments.Researchers from Meta, Rutgers University, Westlake University, and UMass Amherst have developed CLUE (Constitutional MLLM JUdgE), a framework designed to address the shortcomings of traditional image safety systems. CLUE uses Multimodal Large Language Models (MLLMs) to convert subjective safety rules into objective, measurable criteria. Key features of the framework include:Constitution Objectification: Converting subjective safety rules into clear, actionable guidelines for better processing by MLLMs.Rule-Image Relevance Checks: Leveraging CLIP to efficiently filter irrelevant rules by assessing the relevance between images and guidelines.Precondition Extraction: Breaking down complex rules into simplified precondition chains for easier reasoning.Debiased Token Probability Analysis: Mitigating biases caused by language priors and non-central image regions to improve objectivity.Cascaded Reasoning: Employing deeper chain-of-thought reasoning for cases with low confidence to enhance decision-making accuracy.Technical Details and BenefitsThe CLUE framework addresses key challenges associated with MLLMs in image safety. By objectifying safety rules, it replaces ambiguous guidelines with precise criteria, such as specifying should not depict people with visible, bloody injuries indicating imminent death.Relevance scanning using CLIP streamlines the process by removing rules irrelevant to the inspected image, thus reducing computational load. This ensures the framework focuses only on pertinent rules, improving efficiency.The precondition extraction module simplifies complex rules into logical components, enabling MLLMs to reason more effectively. For example, a rule like should not depict any people whose bodies are on fire is decomposed into conditions such as people are visible and bodies are on fire.Debiased token probability analysis is another notable feature. By comparing token probabilities with and without image tokens, biases are identified and minimized. This reduces the likelihood of errors, such as associating background elements with violations.The cascaded reasoning mechanism provides a robust fallback for low-confidence scenarios. Using step-by-step logical reasoning, it ensures accurate assessments, even for borderline cases, while offering detailed justifications for decisions.Experimental Results and InsightsCLUEs effectiveness has been validated through extensive testing on various MLLM architectures, including InternVL2-76B, Qwen2-VL-7B-Instruct, and LLaVA-v1.6-34B. Key findings include:Accuracy and Recall: CLUE achieved 95.9% recall and 94.8% accuracy with InternVL2-76B, outperforming existing methods.Efficiency: The relevance scanning module filtered out 67% of irrelevant rules while retaining 96.6% of ground-truth violated rules, significantly improving computational efficiency.Generalizability: Unlike fine-tuned models, CLUE performed well across diverse safety guidelines, highlighting its scalability.Insights also emphasize the importance of constitution objectification and debiased token probability analysis. Objectified rules achieved a 98.0% accuracy rate compared to 74.0% for their original counterparts, underlining the value of clear and measurable criteria. Similarly, debiasing improved overall judgment accuracy, with an F1-score of 0.879 for the InternVL2-8B-AWQ model.ConclusionCLUE offers a thoughtful and efficient approach to image safety, addressing the limitations of traditional methods by leveraging MLLMs. By transforming subjective rules into objective criteria, filtering irrelevant rules, and utilizing advanced reasoning mechanisms, CLUE provides reliable and scalable solutions for content moderation. Its ability to deliver high accuracy and adaptability makes it a significant advancement in managing the challenges of AI-generated content, paving the way for safer online platforms.Check out the Paper. All credit for this research goes to the researchers of this project. Also,dont forget to follow us onTwitter and join ourTelegram Channel andLinkedIn Group. Dont Forget to join our65k+ ML SubReddit. FREE UPCOMING AI WEBINAR (JAN 15, 2025): Boost LLM Accuracy with Synthetic Data and Evaluation IntelligenceJoin this webinar to gain actionable insights into boosting LLM model performance and accuracy while safeguarding data privacy.The post Meta AI Introduces CLUE (Constitutional MLLM JUdgE): An AI Framework Designed to Address the Shortcomings of Traditional Image Safety Systems appeared first on MarkTechPost.