meta

We thought that the 72 billion that Meta spent on AI in 2025 were outrageous. It was just the appetizer

by

The big tech companies They are spending a lot of money on AI infrastructure and, far from slackening, the figure is only increasing (and not a little). In the case of Meta, the company planned to spend $66 billion in 2025, but in October they had to correct the figure to $72 billion. 2026 has just begun and the figure they have just given is directly insane. Doubling down, literally. Between 115,000 and 135,000 million dollarsthese are the figures that Meta handles for the year 2026. It is almost double what they spent in 2025, a figure that, as we said, they had to correct towards the end of the year, so let’s not rule out that it ends up being more. The spending will mainly be intended to “support the efforts of Meta Superintelligence Labs”, that is, to build more data centers. Which by the way, more than 6 million dollars have been spent on advertising campaigns to convince us that data centers are cool. Record results. The fourth quarter of 2025 has been very good for Meta. Revenue grew 24% compared to the same period of the previous year, reaching 59,000 million dollars and a net profit of 22,800 million dollars, figures that exceed forecasts. According to its CEO, the good results are thanks to the implementation of AI in its advertising services. Recovering confidence. In the previous quarter, the huge spending on AI generated many doubts among investors and, although the results were also good, shares fell up to 8%. This time it has been different and, although spending will skyrocket even more, shares are up 10%. It seems that they trust Zuckerberg’s direction again. Where is the ceiling? It is difficult to know, but what we do know is that since 2023 it has been increasing exaggeratedly. And not a little. According to Meta datain 2023 they spent 28 billion dollars, in 2024 they rose to 39 billion and in 2025 to 72 billion. The jump has been getting higher and higher with each year, I wonder when they will let off the accelerator. All for AI, but without AI. All the big technology companies are spending a lot on AI, but what is striking about the case of Meta is that has not yet launched its new models. To put it in context, Meta’s expected spending is higher than Google’s in 2025, but Google has Gemini while Meta has promises. After the fiasco of Call 4Zuckerberg set himself as a target Completely remodel your AI department. It was spent a fortune in hiring new talentscreated a secret laboratory next to his officeha cut the Metaverse department (also with layoffs) to move resources to AI. The ambition is huge: creating superintelligence. At the moment, what we know is that they are preparing a language model called Avocado and another for image generation that they call Mango. They better measure up, otherwise They will always have the public. Image | Unsplash, Meta In Xataka | Three years after the metaverse fiasco, Zuckerberg has another burning nail for Meta: digital glasses

Data centers are so important that Meta has spent millions on advertising to change our perception of them

by

Meta has spent 6.4 million dollars on an advertising campaign between November and December of last year to convince the American public of the benefits of its data centers, according to the New York Times. The ads, aired in eight state capitals and Washington, DC, featured idealized images of American towns revitalized by these facilities. exists an increasingly significant social rejection on the installation of data centers dedicated to AI, especially due to the impact they have on the excessive consumption of basic resources like light and water. And of course, first we have to convince that they are key so that Meta and the rest of the big technology companies can continue with their operations. The Goal campaign. According to the media, the ads featured emotional stories about Altoona (Iowa) and Los Lunas (New Mexico), two locations where Meta operates data centers. With guitar music and shots of farms and football fields, the videos promised jobs and prosperity. “We are bringing jobs here, for ourselves and for our next generation,” the voiceover said. According to Michael Beach, CEO of Cross Screen Media, Meta “could have purchased these ads with the goal of influencing political decisions and reaching legislators.” Ryan Daniels, spokesperson for Meta, limited himself to say to the NYT that the company pays the full costs of the energy used by its data centers, without commenting on the advertising campaign. Meta is not alone. Just like account NYT, Amazon is funding a similar campaign in Virginia through Virginia Connects, a nonprofit created by the Data Center Coalition. From the Financial Times they point In addition, other operators such as Digital Realty, QTS and NTT Data are also acting more intensely to defend the construction of new facilities. Endurance. In the United States, social rejection has caused the cancellation of multimillion-dollar projects in Oregon, Arizona, Missouri, Indiana and Virginia. Democratic Senator Chris Van Hollen explained He told the NYT that the issue has become “a priority on Capitol Hill” when his voters began to complain en masse about electricity bills. Just like share The media, this month, Van Hollen presented a law to regulate the energy consumption of data centers. Even President Donald Trump spoke out on the matter: “The big tech companies that build them must pay their own way,” wrote a few weeks ago on Truth Social. electricity bill. Data centers have become critical infrastructures for the development of artificial intelligence, but there is increasing social tension over their installation. In October, Bloomberg counted that in the last five years the wholesale price of electricity in areas near large concentrations of data centers in the United States had increased by up to 267%. In Baltimore, residents paid $17 per megawatt-hour in 2020; In 2025 that figure reaches $38. On the other hand, the medium demonstrated In their research, 70% of the points where electricity price increases were recorded were less than 80 kilometers from data centers with significant activity. From Bloomberg they estimate that the energy demand of these facilities in the United States will double by 2035, becoming the largest increase since the 1960s. The situation in Spain. Our country is also experiencing a boom in the construction of data centers. The Community of Madrid, paradoxically the region with the greatest energy deficit in Spainconcentrates a good part of these projects and is expected to reach a power of 1.7 gigawatts in 2030. The consulting firm CBRE pointed out in a report that “there is no investor, operator or large technology company that does not have in its strategic plans to establish its data center project in the Iberian market.” Madrid, together with Barcelona, ​​already competes with cities such as Milan, Zurich or Berlin, although still far from the leading European group in terms of power capacity formed by Frankfurt, London, Amsterdam, Paris and Dublin. What awaits us. According to Bloomberg, the forecasts they point because data centers will consume more than 4% of the world’s electricity in 2035. If these facilities were a country, they would be fourth in energy consumption, only behind China, the United States and India. Meanwhile, big technology companies are already exploring solutions such as modular nuclear reactors (SMR) to power your facilities, or send data centers to space. Cover image | Mark ZuckerbergGoal In Xataka | “The assemblies are not going to be done by AI”: we talk to the kids who have become carpenters, truck drivers and tinkerers

In 2013, WhatsApp cost almost one euro. And nothing prevents Meta from charging for the app again in 2026

by

There was a time when WhatsApp was paid. This was more than ten years ago, before Goal was done with the application and ended up completely changing its structure over time. The latest beta of the app leaks something that seemed inevitable after the arrival of the announcements: a paid subscription to avoid them. He leak. WhatsApp has two versions, the stable and the beta. It is common for the code of the next beta versions to be leaked, giving us a preview of the functions that will end up reaching the final app. And the latest leak points in a very specific direction. Since you recently removed your WhatsApp account from your Accounts Center, the price of your subscription for no ads in Status & Channels has decreased. Review your subscription to accept the new price of %1$s/month; or choose to use Status & Channels free of charge with ads. Additionally, Android Authority has managed to force the code so that the app displays a message in its interface about the possibility of canceling the subscription. WHATSAPP Tricks and tips to HIDE YOURSELF TO THE MAXIMUM and maintain your PRIVACY A plausible hypothesis. So far, practically all the WhatsApp code leaks have ended up materializing: either as functions tested in the beta version, or as features that have ended up reaching the final version. One of them has been the introduction of advertising in the app, which for now is limited to statuses, promotional channels and channel subscriptions. In the case of states, the operation is very similar to what Meta applies on Instagram, interspersing ads every certain number of publications. So… what if I don’t want ads? What do you give me in exchange?. If Meta wants to implement a subscription system with any modicum of success, it will have to offer more than just removing ads in return. The subscription opens the door to new WhatsApp functions, and a business model similar to that of Telegram with its premium version. One in which the app can continue to be used without any inconvenience in the free version, but which opens the door to benefits and a better experience if we check out. Because. If the question is why Meta may intend to charge you for WhatsApp, the answer is very easy: it needs money. In 2014, Facebook paid nearly $1 billion for WhatsApp. Almost 10 years later, He had barely recovered 10% of what he paid for it.. The company continues to need ways to make the investment profitable, and betting on a subscription model is a necessary plan. Image | Xataka Mobile In Xataka | WhatsApp Web: What it is, how it is used and comparison with the mobile app

Meta, Google, TikTok will go to the bench for “addictive design”

by

Today The selection of the jury that will judge Meta, TikTok and YouTube begins in Los Angeles due to childhood addiction to social networks. It is the first time that these technological giants have to defend their business model in court for damages to minors. Why is it important. This is not just another case of inappropriate content or poor moderation. This lawsuit directly attacks the design of the platforms: scroll infinite, autoplay, notifications push and algorithms that maximize screen time. If the plaintiffs win, a precedent is set that could be devastating for the entire industry. The facts. The plaintiff is a 19-year-old girl identified as KGM. She claims to have developed an addiction to networks since she was a teenager. He maintains that the design of these applications was what fueled his depression, anxiety, body dysmorphia and suicidal thoughts. Meta, TikTok and YouTube have denied these accusations and argue that they have invested in security tools. During the six weeks of the trial, Mark Zuckerberg, CEO of Meta, and Adam Mosseri, head of Instagram, will testify. Snap, also initially accused, reached an out-of-court settlement last week for an amount not publicly disclosed. Between the lines. The plaintiffs’ key argument avoids the traditional protection of technology companies: the famous Section 230which exempts them from responsibility for the content uploaded by users. But here the question is not what is published, but rather how the experience was designed to engage minors. The lawsuit openly compares it to slot machines and the tobacco industry: “Defendants deliberately embedded in their products a series of features designed to maximize the engagement youth and increase advertising revenue. The threat. This is just the tip of the iceberg. There are more than 3,000 additional lawsuits in California and 2,000 federal cases pending against these same companies. Several will go to trial this year. The parallels with the trials against tobacco companies in the 90s They are clear and that ended in an agreement of 206,000 million dollars spread over 25 years. A favorable verdict for the plaintiffs would not only cost them billions but would force them to redesign their products practically from scratch, eliminating the addictive mechanics that sustain their spectacular usage figures and therefore their advertising models. The context. Global regulatory pressure has increased greatly in recent years: Australia banned social media for those under 16 in December. France is studying doing the same with those under 15. Other countries such as the United Kingdom and Egypt are currently evaluating similar measures. According to a recent survey by Wall Street Journal71% of Americans would support banning most social networks for those under 16 years of age. Yes, but. The technological they don’t sit idly by: Meta, TikTok and YouTube have launched a public relations offensive by organizing workshops for parents in schools and promoting parental controls. Meta has hired the same lawyers who defended McKesson in the opioid scandal. And TikTok has signed those who represented Activision Blizzard in Previous Lawsuits About Video Game Addiction. At stake. If KGM wins, Section 230 will cease to be the impenetrable shield it has been until now, since it questions how the applications are made, not the content that is uploaded to them. Hopefully this case will end up in the Supreme Court, whatever the verdict. The next six weeks will determine if the scroll infinite and other common practices of these networks have their days numbered, or if there are engagement for a while. In Xataka | An eternally unfocused generation: “I can’t do anything for more than fifteen minutes without looking at my phone” Featured image | Solen Feyissa

Meta was the big loser of the AI ​​race in 2025. She was actually preparing her big move

by

Meta wasn’t dead, but she wasn’t partying either. He was working hard on a new AI model for which there are huge expectations. Now we know a little more about that project, but one thing doesn’t change: it better not fail. what has happened. Andrew Bossworth, CTO of Meta, has confirmed during the World Economic Forum in Davos that the Superintelligence Labs division already has a first internal version of its new AI models. This is an important and long-awaited milestone that they have been working on for six months. “Very good”, but not ready yet. Bossworth did not want to give too many details, but he did indicate that preliminary tests show that the models perform very well. These models will still take time to come to market: Meta is currently in a critical post-training phase for these models to be truly useful for both internal developers and end users. Two great models. Although the names of those models were not specified, rumors and leaks point to two major developments. On the one hand, Avocado, focused on text and which could be launched in the first quarter of 2026. On the other, Mango, focused on image and video generation. A 2025 of transition. The manager described the year 2025 as “tremendously chaotic”, and it was certainly a very complicated year for Meta. He Llama 4 failure It made the company completely change its philosophy and strategy. Zuckerberg did not stop hire talent with a exorbitant costespecially when it acquired Scale AI and signed its CEO, Alexandr Wangnow head of Superintelligence Labs. That investment has also been allocated to acquire companies like Manusthat could become another key component of Meta’s strategy going forward. Glasses as an AI device. If they behave as expected, these models will probably also end up being used in smart glasses from Meta, which has been collaborating with Ray-Ban for years and which you have just seen its second version accompanied, of course, by the striking and even more versatile Meta Ray-Ban Display. Interaction with AI models is one of the most striking features, and these models could take it to new limits. The mystery of Open Source. In July 2024 Mark Zuckerberg indicated that “Open Source AI is the way to go.” Llama was at that time the clear reference, but the disappointing launch of Llama 4 and above all the push for open models from various Chinese companies has made this panorama change significantly. It is not clear that Meta will launch its new models with open versions, and if it did not do so, Chinese hegemony would be even more notable. Will it be worth the investment? Meta is one of the companies that has spent (or bet) the most money on the future of AI. Mark Zuckerberg said that I was willing to lose “hundreds of billions of dollars in AI” because not investing them would be even more dangerous for Meta. He has been consistent with that statement, but It remains to be seen if it ends up working.. The company certainly has the resources to be a protagonist in this market, but today its solutions—with Meta AI at the head—have a very reduced role compared to that of their competitors. Image | Goal In Xataka | China’s best kept weapon in AI is not Qwen: it is the more than 100,000 variables created by other companies

Meta seemed to have more faith than anyone that his metaverse had a future. 1,500 workers have just discovered that they do not

by

In 2021, Zuckerberg was very clear that Facebook’s future was tied to the metaverseso much so that He even changed the name of his company.. However, the market did not respond as expected and, after accumulate million-dollar lossesrecently Meta surrendered to the evidence and put a 30% blow to the budget of the Reality Labs division. It was just the beginning. Layoffs. They overtook him in the New York Times and just confirmed: Meta is going to lay off 10% of the Reality Labs workforce, about 1,500 employees in total. Andrew Bosworth, CTO of the company and head of the division, had summoned employees to the “most important” meeting of the year. So important that for many it has been the last. Cuts. As we said, several weeks ago it was made public that Meta was cutting Reality Labs by 30%. It was an expected decision if we take into account that the division dedicated to the metaverse has accumulated 70 billion dollars in losseswhich is said soon. In this context, the layoffs were the next step and also the confirmation that Meta abandons the dream of the metaverse, at least as they proposed it years ago. New priorities. The objective behind the cuts is to be able to move investment to Zuckerberg’s new “pretty girl”, which is none other than AI. Since the beginning of last summer, Meta has signed big names and AI researchers for real millionaires to create your TBD laboratorywho is engrossed in creation of a superintelligence. In parallel, they are dedicating billions to the construction of data centers, one of them as big as Manhattan Island. They also plan to move resources from the metaverse to the AI glasses, your new reference hardware. Investors have spoken. When Meta announced that it was going to spend even more than planned on AI infrastructure, stocks plummeted even though they had achieved very good results. They were investors sending a clear message: we do not see this unbridled spending at all clearly. However, when the metaverse cuts were announced just the opposite happened and the shares rose. script twist. Meta has not explicitly admitted that it is leaving the metaverse, in fact in October of last year they were still defending it. What they have done is talk about a change in strategy and where before there were VR helmets, now there are AI glasses. It is no longer a virtual world completely separated from the real one, but rather an augmented reality powered by AI. The Ray-Ban Meta they have been a success for the company and recently announced the Ray-Ban Displayalthough We will have to wait to try them. Image | Photo of Azwedo L.LC in Unsplash In Xataka | Meta’s AI director is clear about what generation Z should do: be the future Bill Gates of vibe coding

We will have to wait to test the future with the Meta Ray-Ban Display outside the US: they are being victims of their own success

by

The Meta Ray-Ban Display are not perfect glasses nor do they pretend to be, but it’s hard not to feel that there is something different here. The idea of ​​having a color screen integrated into the lens, with speakers, microphones, cameras and even artificial intelligence, continues to sound more like a prototype than an everyday product. And yet, it is happening. Faced with other attempts to “kill” the smartphone that fell by the wayside, such as the AI Pinthese glasses aim for something more subtle. Screen on the lens, bracelet on the wrist. Beyond the concept, the Meta Ray-Ban Display is based on an unusual combination. The information appears on a small color screen located outside the axis of vision, designed for brief and non-continuous consultations. The control is not done by touching the glasses, but through the Meta Neural Banda bracelet that interprets muscle signals in the wrist area to execute minimal gestures. Meta presents this system as a more natural way to interact, reducing the need for buttons, touch surfaces or visible commands. Overwhelmed demand, expansion on pause. As explained this Tuesday by the companythe volume of interest has far exceeded the available inventory, to the point of generating waiting lists that extend well into 2026. Given this scenario, Meta has chosen to freeze the international deployment that it had planned in the short term and focus all its efforts on fulfilling orders already placed in the United States, while reviewing its availability strategy. This stoppage directly affects the plans communicated months ago. The company had indicated that the Ray-Ban Display would arrive in early 2026 in markets such as the United Kingdom, France, Italy and Canada, a first wave outside the United States that is now on hold. Meta does not speak of cancellation or set new dates, a prudent position that confirms the immediate stop, but does not clarify how long it will be extended. {“videoId”:”x9qouiu”,”autoplay”:false,”title”:”Ray-Ban Meta (Gen 2)”, “tag”:”Meta”, “duration”:”134″} Does the queue get longer here too? Although Spain was not in that first group of countries, the slowdown has obvious consequences for this side of the map. In launches of this type, expansion usually proceeds in waves, starting with the United States and continuing, in some cases, through key markets. In this case, Meta never confirmed plans for Spain, neither for this first generation nor for possible subsequent ones. The only thing that can be stated is that, if at some point these glasses end up reaching this market, the international delay makes it reasonable to think of an even longer wait. In Xataka Meta is so serious about smart glasses that its catalog is already a mess: this is how the new models differentiate themselves News for the Meta Ray-Ban Display. The stoppage in expansion has not prevented Meta from continuing to show progress. During CES, the company presented new functions designed to expand the uses of glasses, as a teleprompter mode to read prepared texts or possibility of writing messages by drawing letters with your finger on any surface, which are then transcribed into digital format. They are improvements that reinforce the idea of ​​a product in continuous evolution, even when its availability remains limited by supply and inventory. Images | Goal In Xataka | Two weeks with the Oakley Meta. Technically impressive, but in no man’s land (function() { window._JS_MODULES = window._JS_MODULES || {}; var headElement = document.getElementsByTagName(‘head’)(0); if (_JS_MODULES.instagram) { var instagramScript = document.createElement(‘script’); instagramScript.src=”https://platform.instagram.com/en_US/embeds.js”; instagramScript.async = true; instagramScript.defer = true; headElement.appendChild(instagramScript); – The news We will have to wait to test the future with the Meta Ray-Ban Display outside the US: they are being victims of their own success was originally published in Xataka by Javier Marquez .

Agentic AI was the new race for Big Tech and Meta was far behind. It has bought the company most capable of recovering

by

Meta has closed the purchase of manusa Singapore-based artificial intelligence startup, for more than $2 billion. Throughout this year, Meta has reinforced its AI operations by acquiring several companies focused on different specialties. In July bought Play AIfocused on voice with AI. In August acquired WaveFormsan audio-focused startup. And in September was done with Rivosa company specialized in the design of semiconductors and RISC-V chips. Manus’s is already the fourth major purchase this year, and it is his hope not to be diluted in the race to dominate AI when all this time he has focused his efforts on Llama and his open weights approach. Why it is important. The Agentic AI (agents capable of performing complex tasks with minimal human supervision) has long become the new battlefield for big technology companies. Although companies like Microsoft or OpenAI had sufficient resources to develop in this field, Meta needed to strengthen its position in this segment if it did not want to be left behind. Manus came to reach 100 million dollars in annual recurring revenue just eight months after its launch, which offers Meta a product that generates money right away, something not very common in this sector. What does Manus do? The startup rose to fame in March with a video demo that went viral, showing how its AI agent was able to produce detailed research reports, build custom web pages, filter job candidates, plan vacations, and analyze investment portfolios. All using AI models developed by companies such as Anthropic and Alibaba. At the time, Manus even claimed to surpass OpenAI’s Deep Research. Currently, the company has around 100 employees, mainly in Singapore, offers subscriptions of $20 to $200 per month and already has a user base of millions. Initial success. Manus emerged a few months after the debut of DeepSeekthe Chinese model that shook the foundations of the industry due to its capabilities supposedly developed with less computing power than its American rivals. Just like account WSJ, the startup secured a $75 million funding round led by Benchmark in April, which valued the company at $500 million. Among its investors are firms such as Tencent, ZhenFund or HSG. Untying ties in China. The parent company behind Manus, Butterfly Effect, was founded in 2022 in Beijing by two Chinese entrepreneurs, including its CEO Xiao Hong, known as ‘Red’. Although most of its researchers and engineers were located in China, Manus launched outside the country because it used American AI models that are not available there. Shortly after securing its investment with Benchmark, the company officially moved its headquarters to Singapore. According to account WSJ, Manus has ruled out developing a version for the Chinese market. Goal declared to Nikkei Asia that, following the acquisition, Manus will have no ties to Chinese investors and will no longer operate in China. All existing investors have been excluded from the operation, according to they count from Bloomberg. What’s coming now? Meta plans to keep Manus running independently while integrating its agents into Facebook, Instagram and WhatsApp, platforms where Meta AI is available. According to WSJManus CEO Xiao Hong will report directly to Javier Olivan, Meta’s chief operating officer. “Joining Meta allows us to build on a stronger, more sustainable foundation without changing how Manus works or how decisions are made,” Xiao stated in the official announcement. No return guarantees. Mark Zuckerberg continues his mission to prove that AI can deliver tangible returns. Goal plans to spend $600 billion in American infrastructure over the next three years, much of it related to AI. Just like assures Bloomberg, it is an amount that causes some skepticism in some investors, since there are no guarantees that this expense will generate significant income soon. Cover image | TechCrunch In Xataka | NVIDIA has paid $20 billion to “license” Groq’s technology. He actually bought it

Researchers extracted photos and statuses from 3.5 billion WhatsApp users. Meta didn’t react until they told him.

by

Between December 2024 and April 2025, a team from the University of Vienna identified 3.5 billion active phone numbers on WhatsApp (practically its entire user base) from a single server and without encountering too much technical resistance. They processed more than a hundred million numbers per hour and extracted not only the existence of accounts, but also public keys, profile photos, status texts, and device metadata. They did it without having to hide, from the same university IP, same server, five accounts. For four months, no one in Meta noticed. Why is it important. This is not the first time that this vulnerability has been demonstrated, as it has already occurred in 2012 and 2021but the first at this scale and speed. The finding exposes a structural contradiction in WhatsApp: Your architecture should show whether a number is registered to enable contact discovery… …but that functional need collides with the privacy of its users. Knowing who uses WhatsApp in countries where it is prohibited, such as China, Burma or North Korea, can have serious consequences. There they detected 2.3 million, 1.6 million and five accounts respectively (not five million, just five). The investigation, published a few weeks ago in NDSS 2026shows that this crack not only persists, but has widened. The context. The researchers developed ‘libphonegen’, a tool that reduces the search space from billions of theoretical combinations of possible mobile phone numbers to “just” 63 billion real candidates for 245 countries. Using unofficial WhatsApp clients that directly access the XMPP API, they queried these numbers at a rate of 7,000 per second. Neither his IP was blocked nor his accounts sanctioned. Meta did not respond until researchers explicitly reported the finding in March of this year, and countermeasures did not arrive until October, just a couple of months ago. The figures. He dataset resulting five times higher the scandal of scraping from Facebook 2021: India leads the document with 749 million users (21% of the total), followed by Indonesia and Brazil. In Spain, 46.5 million accounts. 81% use Android. More than half have a public profile photo. 29% have the status text visible. Between the lines. The researchers were able to infer the operating system by analyzing initialization patterns of the cryptographic keys. Android starts certain identifiers at zero. iOS does this in random values. This detail matters because iPhone users are higher-value targets for attackers. They also detected that public keys are reused. They found 2.3 million different keys used on 2.9 million different devices. In Burma and Nigeria, tens of thousands of numbers shared the same key, pointing either to faulty implementation or outright fraud. They even found twenty American numbers that use a private key composed only of zeros. In detail. The method is not limited to confirming the existence of the accounts. For each one they extracted public keys, timestamps and the list of linked devices. This allows you to build detailed profiles without accessing the content of the messages. The age of the device can be estimated by counting key rotations. The “popularity” of a user is inferred by the frequency of depletion of their prekeys single usewhich are consumed every time you start a new conversation. Researchers downloaded 77 million profile photos of the +1 rank (prefix for the United States and Canada) in a matter of hours. 66% of them contained recognizable faces. They also found disturbing status texts, such as those from traffickers listing prices, accounts business advertising drugs or publicly visible corporate emails from governments and armies. And now what. Meta has deployed probabilistic cardinality counters to limit how many unique accounts a user can query without blocking legitimate contact discovery. It has also restricted bulk access to status photos and texts. The researchers confirmed that the measures work in subsequent tests. But no countermeasures protect those who were already listed during the months in which the system has been wide open. The big question. For four months, from a university server without even hiding their identity, they looted practically the entire user base of the most used application on the planet without anyone at Meta realizing until they were explicitly told. If these researchers were able to do it under these conditions, who else did it before without telling anyone? In Xataka | WhatsApp brings the big update of the season: the most important change is not on the mobile, but on the computer Featured image | Dimitri Karastelev

Meta has been bragging about LLaMa for years while missing the AI ​​party. And she’s already tired of being the Android of AI

by

Close your eyes and think about AI. It’s easy for the names that come to mind to be ChatGPT either Geminiand it makes perfect sense: OpenAI and Google have focused on pushing solutions for real users. The one that may sound familiar to you, but you don’t even remember, is LLaMa. Meta has focused for years on AI for the sector, forgetting the consumer. And that’s about to change with Mango and Avocado. Because the “new” Meta no longer wants to be the Android of AI: it wants to embrace the Apple model. The ‘meh’ of LLaMa 4. Meta’s approach to artificial intelligence has been, and is being, curious. LLaMa 4 It was a frustrating release, one that hasn’t lived up to expectations. It competed with GPT-4 (let’s go for 5, whose launch also brought controversy), but while OpenAI and Google have struggled to position their AI models as options open to the user thanks to the chatbot, Meta has gone in other directions. They have their Meta AIbut LLaMa was the star product. They have ‘gone’ from the user and have focused on professional options. Meta opted for Open Source (in quotes) seeking to turn LLaMa into the foundation on which everything that has to do with AI is built. To make a simile, Meta wanted LLaMa to be the “Android of AI.” It hasn’t worked out, and now it wants to pivot to an Apple-style model: closed and consumer-oriented. 14.3 billion dollars. That’s the money that Zuckerberg, in full ‘founders mode’ like Florentino Pérez has been left in Scale AI. The startup has established itself in a very short time as the great promise of general AIone of the short-term objectives of the majors in the sector. And, now, it is owned by Meta. It is the Madrid of the galactics. Because even if Scale AI did not develop ChatGPT, Gemini or Claudehas built the infrastructure for proposals of this style. And with the purchase comes Alexandr Wang, who was CEO of Scale AI and now director of AI at Meta. It seems that the relationship is already bearing fruit. Mango and Avocado. As we read in WSJWang has mentioned two new AI models that will go live in early 2026. And the proposal is radical considering where the company came from: Avocado – This will be the brain and successor of Calls. It is scheduled for the first half of next year and is expected to be the one that begins the new era of privatization of the model: it would mark the transition to a closed system. Mango – If Avocado aims to be invisible to the user, Mango will be the complete opposite. It will be an image and video generation model to compete directly against sora, I see either Nano Banana from OpenAI and Google, respectively. Less papers, more chats. Thus, Zuckerberg and Wang will be able to have a model that people associate as synonymous with “artificial intelligence.” Google and OpenAI have come a long way, but if AI has taught us anything, it is that new tools can become popular in a heartbeat if they hit the right button. Midjourney was the grail of generative AIFor example… But of course, neither Google nor OpenAI are going to sit idly by. Both are burning money to continue developing their models and the problem is that, although what they get Meta works like magicthey will arrive years late consumer AI competition. They have dispersed their AI in WhatsApp, Instagram and professionals instead of having a single chatbot; have published studies on how capable your artificial intelligence is. In the middle of all that, they are late to the party. And, precisely, in that They look a lot like Apple. Images | Mark Zuckerberg, Dima Solomin In Xataka | “AI is unstoppable”: the CEO of Freepik talks to us about AI, entrepreneurship and the mistakes of an EU that only focuses on the dangers of AI

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.