question

Germany is trying to stop its electricity dependence on China. The question is whether that is even possible.

by

Almost four years ago, Germany learned a painful lesson: your industry cannot depend on the energy of a geopolitical rival. The Russian gas crisis after the invasion of Ukraine forced the Germans to make more than one sacrifice while the country’s energy model was transformed. Now, at the gates of 2026, Friedrich Merz’s government faces a déjà vu disturbing. The same stone twice. Germany may have become independent of Gazprom’s gas pipelines, but its solar panels and grid technology bear, directly or indirectly, China’s stamp. Good: Berlin has just hit the brakes. The collapse of a seemingly innocuous financial operation last week has revealed that Germany is carefully reviewing every watt that enters its system to avoid repeating the historic Russian gas mistake. The trigger. The Italian company Snam SpA intended to acquire a minority stake in Open Grid Europe (OGE), one of the largest gas network operators in Germany. On paper, it was an investment between European partners. In practice, the German Economy Ministry saw the shadow of Beijing. The problem was not Snam, but its shareholders. The state-owned State Grid Corporation of China owns 35% of Cassa Depositi e Prestiti, which in turn owns a third of Snam. For the Merz government, that was risk enough. Given Berlin’s refusal to accept the proposed solutions, Snam withdrew its offer last week. A clear message. Berlin does not want companies with Chinese state participation to have access to the country’s energy arteries, even indirectly, which marks a change in doctrine compared to the era of Olaf Scholz, who at the time allowed the Chinese shipping company Cosco to enter the port of Hamburg. The current executive is much more defensive: national security takes precedence over capital. The question is… Too late? If blocking the purchase of a gas network is relatively simple, unraveling technological dependence on China is a logistical and economic nightmare. 95% of the photovoltaic cells installed in Germany come from Chinese manufacturers. And almost the entire wind industry, especially offshore, depends on rare earths controlled by China. The German energy transition is based on Asian hardware. Germany needs Chinese technology to meet its climate goals. And he doesn’t hide it. The German government has already raised this concern in international forums, denouncing the Chinese overcapacity in sectors such as electric mobility and solar energy. Technology that is needed but now considered a “systemic risk.” Is decoupling possible? In 2018, the German government already had to intervene so that the state bank KfW bought a stake in the network operator 50Hertz, preventing it from falling into the hands, again, of the Chinese State Grid. Seven years later, the strategy of “patching” individual acquisitions seems insufficient in the face of structural dependence. If the experience with Russia is any guide, Berlin seems to have decided that, this time, the price of security must be paid in advance, before anyone decides to turn off the tap. But today, the reality of the market is stubborn: replacing Chinese hardware means, almost invariably, paying more and taking longer to deploy renewables. Image | rawpixel In Xataka | If you were expecting cheap electricity this winter, we have bad news: Holland

OpenAI just launched ChatGPT for teachers. The question now is how much education we are willing to delegate to AI

by

What happens when a teacher uses artificial intelligence to prepare his classes, a student uses it to do homework, and finally, that same teacher uses AI again to correct them? It may not be the norm yetbut that scenario no longer sounds so far away. The speed at which these tools have been integrated into classrooms has opened a fundamental debate: what do we really learn if we let technology do the work for us? And what does the educational system lose if this process becomes a habit? The landing of AI in education is neither coincidental nor recent. Technological tools have been present in classrooms for years, with platforms such as Google Classroom either Moodle. The novelty is not in using technology, but in relying on systems capable of generating content, proposing solutions or even being used in pedagogical decisions. That is where the big developers—Google, Microsoft, Anthropic and, more recently, OpenAI—have decided to go a step further and position themselves at the center of the educational debate. Here OpenAI lands with a dedicated proposal for teachers in the United States. We are talking about a version of ChatGPT Designed for primary and secondary educators, free for verified teachers, with administrative controls for centers and school districts. Unlike the service that almost all of us know, OpenAI ensures that the data generated in these environments will not be used, by default, to train its models. What ChatGPT offers for teachers Personalized assistance. It allows you to enter school level, curriculum and desired format so that the answers adapt to the real style of the classroom. It is the teacher who controls that configuration. Integration with usual resources. You can generate presentations with Canva, import lesson plans or documents from Google Drive and Microsoft 365, and start a conversation with that context already activated. Ideas from other teachers. Show real examples of teachers already using ChatGPT in their classes, directly below the editor, as a source of inspiration. Teaching collaboration. It makes it easy to create custom GPTs and shared templates to plan units, lessons, or assessments among colleagues in the same school or district. Management from the center. It offers a manageable workspace, with secure accounts and differentiated roles for teachers and academic leaders. What is OpenAI pursuing with this? Among the 800 million weekly ChatGPT users there are many teachers. The company explains that they are using the tool to design teaching units, adapt the curriculum to regional standards or generate examples that help evaluate their students. Let’s look at some of the usage examples you have shared: Generate examples for a task You are an expert English teacher. Using the prompts in the accompanying readings, generate seven different sample answers. Responses should be one paragraph in length and range in quality from very well written to very poor. They must be written following the RACES format (restate, respond, cite, explain and summarize). Include a justification for each answer, indicating your level of writing. Plan a multi-week drive My science department is redesigning the 8th grade physical science curriculum and I need help creating a teaching unit based on the attached objectives. Please make a plan for a 20-day unit with 55-minute classes. I need a guiding question for each day to help focus learning. Provide hands-on activities for students to explore these topics. As we can see, AI is here to stay, and trying to ignore it is not an option. The real question is how to use it without replacing the act of learning, which is much more than completing a task. Because if the teacher uses AI to solve what he has to prepare, and the student does the same to deliver what is required of him, what remains of that process beyond compliance? The educational system is not based on the ability to deliver results, but on the ability to think, make mistakes and argue with one’s own knowledge. An MIT study provides data that begins to illuminate the debate: users who wrote essays with ChatGPT produced the text 60% faster, but their cognitive effort was relevant was reduced by 32%. That is, they achieve a more polished result, but with less mental work. Another study, in this case from the SBS Swiss Business Schoolnotes that the increased use of AI is linked to the deterioration of critical thinking skills. We still do not know what effects this dynamic will have in the medium or long term. What we do know is that the classroom has become a territory where big technology companies want to be. And that the real educational challenge of the next decade will not be deciding whether we use AI, but deciding how much of the educational process we are willing to delegate to it. Images | Xataka with Gemini 3 | OpenAI In Xataka | The problem is not that the AI ​​is not able to read the time. The problem is confirming that he does not reason and only repeats what he has seen.

The only thing that Europe’s AI Law has achieved is to leave us lame. The question is whether turning back will do any good.

by

December 8 was a fateful day for the European Union, but not many realized it. And it was because that day the AI ​​Act was passedthe European regulation on artificial intelligence. Thierry Breton, European commissioner, he was pleased with a tweet that automatically became a meme. I was bragging about how Europe had tripped itself up. The responses to that tweet They made it clear that the reception of the regulations was very different from what the EU would have expected. The criticism was forceful and very clear: with these regulations the only thing the EU was achieving was to slow down innovation and make it even more difficult to compete in a segment that was defining the world. While the US and China joined the party without asking permission and without asking for forgiveness, Europe stayed at home happily crocheting. That regulation, which came into force in August 2024instantly caused the AI ​​segment out at two speeds: that of Europe, almost at a standstill, and that of the rest of the world, which stepped on the accelerator (without looking too closely at the consequences). We have seen the consequences of that in the last two years. Europe has been relegated to the second (or third) plane, and with honorable exceptions like the Spanish Freepik or the French Mistral, we have very little to talk about in this area. Meanwhile, the US dominates the commercial plane and China is a steamroller both at a training level as in your open model development. Europe wants to turn back: the question is whether it is too late Yesterday the European Commission presented a project for simplify various digital regulationsand the most important modifications actually affect the General Data Protection Regulation (GDPRor GRPD for its acronym in English). The changes proposed by the Commission will make it easier for companies to share sets of anonymised and pseudo-anonymised personal data. That will have a direct impact on the capacity of AI companies, which They will be able to legally use personal data to train their data models as long as that process meets the rest of the GDPR requirements. The proposal also softens one of the key elements of the AI ​​Act, which, as we say, came into force in August 2024 but included several elements that would come into force some time later. Thus, now the “grace period” for the regulations that regulate the high risk AI systems —those that pose a “serious risk” to health, safety or fundamental rights—is widespread. It was supposed to be activated in summer 2016, but now that regulation will only apply when it is confirmed that “the necessary standards and supporting tools are available” for AI companies… whatever those standards and tools are, yet to be defined. Other amendments in that new Digital Omnibus include simplified requirements for the documentation required of SMEsin addition to a unified interface so that companies can report cybersecurity incidents. Henna Virkkunen, vice president of technological sovereignty at the European Commission, explained that: “In the EU we have all the ingredients to be successful. However, our businesses, especially startups and small businesses, are often held back by a set of rigid rules. By reducing bureaucracy, simplifying EU legislation, opening access to data and introducing a common European business portfolio, we are creating space for innovation to be produced and commercialized in Europe. This is being done the European way: by ensuring that users’ fundamental rights remain fully protected.” These amendments to current digital regulations will now have to be approved by the European Parliament and the 27 member states of the European Union — which will need a qualified majority— to approve it. That process could last months, and during it the proposals themselves could see notable changes before being applied. As indicated in The Guardianthis “massive setback” of this regulation has caused concern among groups fighting to continue protecting privacy of European citizens. The European Digital Rights (EDRi), a pan-European network of NGOs, Indian that if the changes to the regulation are accepted, it will become easier for technology companies to collect and use personal data to train AI models without asking for consent. The European agenda seemed to change when former Italian Prime Minister Mario Draghi warned last fall of how Europe had fallen worryingly behind in the technology race. That speech was a breath of fresh air for Europeand European business groups have welcomed the proposal with optimism, but believe that they still fall short. A representative of the Computer and Communications Industry Association of which Amazon, Apple, Google and Meta are members indicated that “efforts to simplify digital and technology regulations should not stop there.” One click for cookies This simplification of regulation that affects all types of digital scenarios can have a positive effect. Accepting or rejecting cookies has become a daily torture for millions of Europeansbut the user experience may improve significantly in the coming months. And it may get better because the EU has proposed a modernization of policies related to cookies. To try to improve the browsing experience, it will limit the number of times cookie warning banners appear, but also will make it possible for us to accept or reject cookies with a single click. In fact, the future may be even more promising, because what is intended is that said consent (or denial) of cookies is integrated into our browser so that once we configure it, the websites are not constantly asking us if we accept cookies or not: the browser will know what we want and will answer for us at all times. In that “digital package” it is specified that once we accept or reject cookies with that “single-click“, websites must respect that choice of citizens for six months. Image | Christian Lue In Xataka | For the EU, our privacy has always been more important than AI. Until he understood that he was left behind

Gemini 3 promises more quality and precision than ever in its responses. The question is whether we will really notice the difference

by

Google has announced the launch of Gemini 3its new artificial intelligence model. in the company They claim it is their most advanced reasoning model because it is “designed to understand depth and nuance.” Gemini 3 will also be available as standard as part of AI Mode in the renewed Google search engine (in this case and for the moment, only in the US). It is the first time that Google offers the benefits of its AI model from day one in the search engine, but it also reaches the Gemini app and the developers who work with AI Studio and Vertex AI. Behind him success of Gemini 2.5 Pro and Flashthe new version arrives in 30 new languages, including Catalan, Basque and Galicianand as we say you can start testing today in the United States… or outside of there via a VPN. Gemini 3 promises. At least in the tests Google highlights how the model’s behavior has been outstanding in various synthetic tests. Thus, Gemini 3 leads the LMArena classification with 1,501 points—the first to overcome the 1,500-point barrier. According to Google, the Gemini 3’s test results put it ahead of all its competitors in virtually all scenarios. In fact, he manages to reason “at the level of a PhD” according to the tests of Humanity’s Last Exam (exceeds 37.5% of the test without tools) and GPQA Diamond (91.9%). It also makes spectacular progress in mathematics, as demonstrated by the 23.4% on the MathArena Apex test: GPT 5.1 scores 1.0% and Claude Sonnet 4.5 1.6% on the same test, for example. The model also wants to be more direct: his answers are more “concise (…) and he prefers to offer valuable information instead of resorting to clichés and flattery. Tells you what you need to hear, not just what you want to hear“. Gemini 3’s ‘Deep Think’ mode goes even further in tests: in Humanity’s Last Exam it achieves 41.0%, but it also in the demanding ARC-AGI 2 It achieves 45.1% (with code execution), which also demonstrates progress in abstract reasoning and visual understanding. Gemini 3 explains the world to you in a simple way The model has a context window of up to one million tokens, which allows it to be used, for example, to analyze huge repositories of code or text and then work on that data. Its multimodal support allows you to analyze all types of information. For example, Gemini 3 can decipher and translate handwritten recipes in different languages ​​to create a family cookbook that you can share. Or analyze your pickleball games (we assume the same thing happens in other sports) and identify areas where you can improve and generate a training plan. Or scrutinize the data from a research paper and from it generate code for an interactive guide that helps us better understand those studies. In fact, integration with Google Search is an especially important part of Gemini 3, which being “embedded” in AI Mode It has the capacity to generate interactive visual elements (widgets, calculators, simulations) in real time. At Google they want the search to be more interactive than ever, and that will mean that sometimes the answers will not be just text, but rather a small interactive webapp that allows us to better understand the answer. Programming (and agents) to power The other crucial element of the model is its capacity in the area of ​​programming. Its results in tests of this type are once again outstanding, and for example it tops the WebDev Arena leaderboard with a score of 1,487 ELO. The model now behaves much more powerfully in the visual part. It also scores 54.2% on Terminal-Bench 2.0, which evaluates a model’s ability to use tools and operate a computer through a terminal. Additionally, it far outperforms 2.5 Pro in SWE-bench Verified (76.2%), a benchmark test that measures the effectiveness of scheduling agents. These Gemini 3 programming capabilities are intended to be used in a new agent development platform called Google Antigravity. The developer experience is using a “conventional” AI integrated development environment (IDE), but your agents can have access to the editor, terminal, and browser. That means these agents can autonomously plan and execute complex software tasks and validate their own code, making it easier for human developers to review and audit that code than ever before. The real challenge of the most recent models On paper Gemini 3 is postulated as a model that can really make a difference compared to its competitors. The test results and Gemini’s own trajectory make us think that the behavior of this model will indeed be remarkable. However The question is whether we will really notice the difference. In recent months we have seen how other AI companies have launched new models, but the impact for a large majority of users has been discreet: the previous models already performed really well, and although the new ones undoubtedly provide improvements, for many consultations these improvements allow us to perceive that jump in performance. Here we see two ways for Google to effectively demonstrate the capabilities of these models. The first opportunity for Gemini 3 will likely be in the area of ​​programming, and it will be these professionals who will likely be able to get the most out of those additional capabilities. But for the rest of the users, it will be that new AI Mode and the Gemini app that will have to make us notice those features. We are intrigued by this ability to respond with small interactive elements —graphics, widgets—, and perhaps with them we will really discover this new capacity of this chatbot. In Xataka | Let’s say goodbye to Google Assistant a decade later. Google has begun to delete its code to leave only one option: Gemini

If the question is why are non-alcoholic drinks so expensive if they are not taxed, the answer is simple

by

Taking a look at the drinks menu of any establishment is a contradiction: non-alcoholic beer It is worth the same as one with alcohol. The same thing happens as with the decaffeinated coffee and the easiest thing is to think that it doesn’t make sense. If you don’t have alcohol, the rules don’t apply. specific taxes on alcohol. The problem is that there are a lot of factors that come into play. The contradiction. Than the price of non-alcoholic beer equal The counterpart with alcohol is something that is not reserved for locals: it is also seen on supermarket shelves. The price of these versions not only equals that of alcoholic beverages, but can exceed it in some cases, and is not limited to beer: also non-alcoholic wine or to refined alcohol products. It’s… strange, especially considering that there are a series of taxes levied on alcoholic products. Guardian echoed this situation, pointing out that the prices of a liter of non-alcoholic beer It is 5% higher than the alcoholic counterpart in supermarkets, 25% higher in pubs. Cider without is 10% more expensive than with and with wine and liquors Something curious was happening: the same price or cheaper in the supermarket, more expensive in the bars. Taxes. In the United Kingdom, about 10% of the price of beer are taxes, but it is not something exclusive to the islands. In Spain, Italy or France there is also the tax to beer and it depends on whether they have more or less alcohol, also if it is artisanal or not. Wine has VAT in Italy, Germany and Spain, but in France it has a tax between 4 and 10 euros per hectoliter and the highest taxes are observed for distillates. That is to say, it is evident that part of what is paid for a non-alcoholic drink is taxes and logic tells us that, if a drink does not have alcohol, it should be between a little cheaper -beer- and much cheaper -0% spirits-. The reason why this is not the case is quite simple. R&D. There are three elements that come into play to prevent it from happening. The first is that, in many cases, production is more complex and expensive than that of alcoholic beverages. In the case of non-alcoholic beer and wine, production starts exactly the same as with alcoholic versions. This implies that the drink is made with fermentationwhich is what raises the graduation. However, then you have to take that extra step that costs money: dealcoholization. It is something that involves specific technology to remove alcoholic content preserving both flavor and texture. In the elimination process, part of the liquid is lost, so producers must use more raw materials to “fill” and, in addition, the alcohol works as a flavor enhancer and, when eliminating it, it is necessary to incorporate additional ingredients such as extracts, aromas or whatever each brand has in its formula. In short: it is not so much the ingredients as the times and processes, which are not eliminated with alcohol, but rather increased. “The industry has made the decision that non-alcoholic drinks are versions of premium products, seeking to ensure that ‘non-alcoholic beer’ is not associated with something cheap and of lower quality” Economy of scale. More or less. That is one of the factors. The second is that yes, it seems that we have embarked on the fashion to stop consuming so many alcoholic beverages. It is something that the industry, especially the beer and wine industry, has observed in recent years, when there has been a significant increase in consumers of non-alcoholic products. If we look back, the non-alcoholic beer market has explodedbut if we look at the total, non-alcoholic beverages only represent a small percentage of volume sales in the alcoholic beverage market. Since there is less demand than the counterpart with alcohol, they do not benefit from economies of scale. That is: the factories that produce bottles, cans, labels, advertising and the alcohol products themselves produce such a high quantity that the cost per unit is low. When non-alcoholic drinks are produced, different labels are made, but as the quantity produced is smaller, the cost per unit is higher. As for the big brands: the independent ones that only produce non-alcoholic drinks have invested a lot of money in research and machinery and cannot afford aggressive margins because they want to recover that investment. and psychology. And the third factor is something that seems silly, but also plays an important role in all of this. The Guardian article alluded to the fact that wine or non-alcoholic spirits were priced the same or lower than alcoholic versions in the supermarket, but in bars, things were different. And it is something that has to do with the positioning of the brands and the perception of the user themselves. Mixing the psychology and marketingif the price of one of the products were significantly lower, it could be perceived as inferior quality. Therefore, in the case of beer, for 0.0 to be seen as a legitimate substitute, the price must be comparable to the alcoholic equivalent. If we see a price equal to or slightly lower than the alcoholic equivalent, the reason may be that it is a version made by an already established brand, with a massive infrastructure that allows them to play with margins and their own brand image. And it also comes into play that non-alcoholic beers from not so long ago were pretty bad. They have improved a lot in recent years, but John Holmes, director of Sheffield Addictions Research Group (a public health think tank based at the University of Sheffield), point that, to improve the image, “the industry has made the decision that non-alcoholic drinks are versions of premium products, seeking to ensure that ‘non-alcoholic beer’ is not associated with something cheap and of lower quality.” He assures that “if you want to reform the reputation of a product, you launch a premium version.” … Read more

There was a time when HTC sold more phones than Apple and Samsung. The question is what happened next: Crossover 1×28

by

In 2002 we still didn’t have smartphones, but I was lucky enough to see a preview of that future. I traveled to London with Microsoft and at that event the company presented the Orange SPVa big-headed and different mobile because it was based on Windows Mobile 2002. In it you could surf the Internet, write emails or listen to music, although in a limited way because neither the software nor the hardware were very competitive at that time. And yet, the vision was clear: everything was going toward those devices. What was surprising was not only that, but who manufactured that device was HTC. The Taiwanese firm was already beginning to be known for manufacturing devices for others, but it would soon end up launching into the smartphone market taking advantage of the push of Android. In 2011 its market share in the US became superior to Apple’s or Samsung, but after that achievement, the firm started making bad decisionsand other manufacturers joined in – especially from China – who began to make competition much more difficult. HTC never recovered from that and although it experimented with other segments like virtual realityfaded to a paper totally secondary in the technological field. We talk about all this in a new episode of Crossover in which we remember the great milestones of the company and that singular fall almost into oblivion. In Xataka | “It is a brutal economic effort, but we have to act now”: parents who are taking their children to schools without screens

If the question is whether you have to pay garbage tax for a parking space in Madrid, the answer is: good luck with the Cadastre

by

April 8, 2022. The Government publishes in the BOE Law 7/2022, on waste and contaminated soils for a circular economy. Behind this name hides a small bomb that has been exploding, little by little, in each municipality. In Madrid, that detonation has come this year. Beyond the calculation, there are thousands of car parks that are now wondering: do I have to pay the new garbage fee? Where do we come from? My colleague Carlos Prego explained it a few days ago in Xataka. Madrid has recalculated its garbage rate, making reference to the famous Law mentioned above with a calculation that the OCU has come to define as “original and unfair”. The point is that controversy has arisen because Madrid City Council said “eliminate” this rate in 2015, alleging that they removed the tax burden from the citizen. The 2022 Law obliges municipalities with more than 5,000 inhabitants to begin collecting it, following European guidelines. To calculate that rate, The City Council has taken into account the cadastral value of the apartments or the tonnage of garbage that is collected in each neighborhood. That is, those who live in a neighborhood where more garbage is generated will pay more… and that directly affects neighborhoods with great tourist activity (hotels, tourist apartments…), commercial or very densely populated. a truce. The criticism has been so virulent on the part of the oppositionof the neighbors and of the associations of consumers who the City Council has partially rectified. They assure that now it will be taken into account the number of registered in each household looking ahead to next year. But what happens where no one lives? Yes, where, for example, there is a parked car because we are talking about a garage. And the garbage rate also affects the owners of a parking space… At least, apart from them. and a battle. Because although the neighbors seem to have received a truce with the new calculation in the garbage rate, which, yes, the City Council continues to defend that it will have little impact on obvious changes for neighborsthe new open front is what happens to the parking lots. And the door had been opened for a neighbor to have to pay a garbage fee for his home and another garbage fee for his parking lot. Despite the fact that, obviously, the garbage generated by a parking space is minimal or non-existent. Little more than general cleaning if we talk about a community parking lot. However, the rate taxes the provision of the service of collection, transportation and treatment of urban waste, in the words of the College of Administrators. That is, the same person (house and garage) could be charged for a single garbage collection. Who pays then? Those who will pay. Those owners of parking spaces whose parking lot is registered in the Cadastre as a “parking-industrial-use warehouse”, in the words of a circular sent by the Madrid College of Administrators to the Property Administrators of the Capital. What does this mean? They clarify it from the Cadastre which, upon consultation with one of these administrators, have confirmed that they are those independent garages that cannot be accessed from a home or from the common areas of a building. That is, those in which garbage is collected individually. Those who will not pay. Those owners of a parking space whose parking is registered in the Cadastre as “residential use”. Or, in a simplified way by this last entity, which are accessed from a home or from common areas with another building. In that case, they may be communities of different owners (garage and building) but if access is from the same common areas, the former will not pay the garbage fee. What does the City Council say? That they adhere to the type of land use specified in the Cadastre and, therefore, that it is the latter that specifies who should or should not pay the garbage rate. The only solution given in this case by the College of Property Administrators of Madrid is for the community to present a declaration of cadastral alteration to specify that the land use is residential and does not correspond to industrial use. The other alternative is to present a written due to discrepancies with the description of cadastral use. Photo | Kertis Stick and Madrid City Council In Xataka | The best horror movie of this winter has been released. And the protagonists are the owners of a home in Spain

171 million euros later, Metro de Madrid wants to reopen line 7B. The big question is whether the tenth time will be the charm.

by

Line 7B of the Madrid Metro will fully reopen this same month of November after more than three years closed. It is the tenth attempt to normalize a service that was inaugurated in 2007 and that has accumulated more than 800 days without functioning since then. The total cost of repairs reaches 171 million eurosnot counting compensation to neighbors, which already exceeds 23 million and continues to increase. A disaster that began in 2007. When Esperanza Aguirre promoted this expansion to have it ready before the regional elections of 2007, no one could imagine the consequences. The construction of the tunnel seriously altered the subsoil by bringing salt and water into contact, which caused the progressive dissolution of the soil. The result: collapse of the tunnels, massive water leaks and structural damage to hundreds of homes in San Fernando de Henares and Coslada. According to internal documents obtained by El Paísalready in 2008 the technicians warned of the “risk of collapses in the metro tunnel and the surrounding buildings”, and in 2009 they warned that action was “extremely urgent.” The figures of the disaster. The repair bill includes 117 million invested by the Ministry of Transport in works and compensation, 49.7 million from the Canal de Isabel II in hydraulic infrastructure, 2.4 million from the Metro itself and 1.7 million from the Ministry of Education to demolish the El Pilar educational complex. In total, more than 171 million euros. But the number will continue to grow: Property compensation, which in 2022 was estimated at 12 million, has already reached 23.3 million and there are nearly 300 open files. Additionally, 73 homes had to be completely demolished, leaving families paying mortgages on homes that no longer existed. The technical solution. To stabilize the ground, the Community has injected more than 11,000 tons of mortar of concrete in the subsoil through 26,000 drillings that reach up to 45 meters deep. It has also deployed 179 mini topographic prisms inside the metro and laser sensors that send daily data on ground movements. The Polytechnic University of Madrid analyzes also satellite images to detect any anomaly. According to the Minister of Housing, Transport and Infrastructure, Jorge Rodrigo, 511 surveillance elements and five robotic stations have been installed that will constantly monitor the road, the land and nearby buildings. The neighbors don’t forget. Although the Community assures that the infrastructure now presents “stability” and meets “the necessary security conditions”, those affected they maintain their mobilizations and demand greater compensation in court. Furthermore, a study by the Polytechnic University detected “considerable movements” in distant areas “without stabilizing”, although without specifying more details. For the 120,000 inhabitants of San Fernando de Henares and Coslada, the November reopening is just the first step to move forward in almost two decades of nightmare. And now what. The Community will allocate an additional 8.2 million to surveillance and maintenance contracts to act immediately in the event of any incident without the need for emergency contracts. Line 7B will be the most monitored infrastructure of the Madrid Metro, precisely because it is the one that has caused the most problems. It remains to be seen if this time the line is truly stable or if it will close again, as has happened on nine previous occasions. Cover image | Zarateman (Wikipedia) In Xataka | Madrid and Lisbon will be linked by the AVE. It will only arrive (if it arrives) 24 years late

If the question is how spacious the Starship will be, the answer is yes

by

Last week, NASA’s acting administrator proposed study alternatives to SpaceX’s Starship to send astronauts to the Moon before China does. SpaceX has just published a blunt response. A paradigm shift. The self-imposed moon race against China has made the United States forget the real reason why NASA chose the SpaceX’s gigantic Starship for his return to the Moon. As SpaceX itself has been responsible for remembering in a long publication Loaded with images, technical details and advances that we were unaware of, its Starship HLS (Human Landing System) is not a lunar landing module like that of the Apollo missions: it is a paradigm shift designed to build a permanent lunar base. Size comparison between Starship HLS and the Apollo lunar module This is Starship HLS. The comparison is almost comical. While the Apollo lunar module, that took the first humans to the Moonmeasured seven meters high, Starship HLS will rise vertically to 52 meters. To put it in terms of room to stretch your legs: The Apollo lunar module had the habitable volume of a wardrobe (4.5 cubic meters). The Lanyue spacecraft that the Chinese astronauts will use has twice the volume. Starship, according to SpaceX itself, will have two-thirds of the pressurized volume of the entire International Space Station (613 cubic meters). What’s more, the SpaceX ship will have two airlocks for exits to the surface. Each of them will have a habitable volume of 13 cubic meters, which means that a single Starship airlock is more spacious than the Chinese lunar landing module that NASA is so concerned about. Render of the Starship HLS cone inside A luxury apartment. If the size comparison wasn’t enough, new renders of Starship’s interior show a level of comfort that no spaceship has ever had outside of sci-fi movies. Forget the image of astronauts crammed into an aluminum can. What we see is a spacious, multi-story interior, with a clean and futuristic aesthetic. There is a spiral staircase, a control area with multiple seats and a bay window offering panoramic views of the lunar surface. Astronauts inside Starship HLS A beast of burden. Starship is not designed to carry new American flags to the Moon. As SpaceX has taken care to remember, it is designed to fulfill the initial promise of the NASA’s Artemis program: create a “permanent and sustainable presence on the Moon”, building a lunar base. Starship cargo variants will be able to land up to 100 tons directly on the lunar surface. This includes pressurized and non-pressurized rovers, nuclear reactors for power generation like the one NASA wants to install before China, and prefabricated lunar habitats. 2026 will be the moment of truth. SpaceX says it has completed 49 key milestones in Starship’s development, including demonstrations of life support systems, testing of landing legs, qualification of the docking adapter, and demonstrations of the elevator and airlock. However, the big obstacle remains refueling ships in orbit to compensate for the evaporation of cryogenic fuel, something that SpaceX hopes to achieve in 2026 with the new Starship V3. Without fuel transfer in orbit, Starship cannot reach the Moon with its crew and its 100 tons of cargo. Images | SpaceX In Xataka | The enormous size of Starship, in images that give an idea of ​​its scale In Xataka | A genius named Tom Mueller designed the engines for the Falcon 9. And now that genius wants to beat SpaceX on its own turf

Anthropic is spending much more money than it brings in. The question is how long can it continue like this?

by

How much does AI cost? That question can be answered by AWS, which has billed Anthropic a whopping $2.66 billion so far this year. The problem is twofold, because in that same period it is estimated that Anthropic has earned 2.55 billion dollars, so with that alone it has spent more than it earns. But Anthropic has many more expenses and the accounts, once again, do not work out in the AI ​​segment. Why is it important. The data revealed by Ed Zitron confirms the problem they face all AI startups: They spend (much) more than they earn, and that trend does not seem to be reversing. In fact, although these companies are growing in revenue, they are also growing proportionally in expenses. And the question, of course, is whether this pace is sustainable. The Anthropic case. According to Zitron data, in 2024 Anthropic earned between $400 and $600 million, but spent $1.35 billion on AWS, that is, 226% of its income. The trend appears to continue in 2025, because the share of spending on AWS is 104% of its revenue. It seems that things have improved, but that expense does not include what it costs Anthropic use Google Cloud infrastructureanother of its partners in all its operations. The expenditure on it is also likely to be enormous, which complicates the situation. The mystery of unexplained costs. The unaccounted cost gap is also enormous. In 2024 Anthropic’s total spending was estimated at 6.2 billion dollars. If we know that he spent $1.35 billion on AWS, there is $4.85 billion left that is not explained. That suggests that spending on Google Cloud and other operational costs is absolutely astronomical. In fact, computing costs may be much higher than we thought. Another startup desperate for investment. Meanwhile, Anthropic continues to raise capital. Zitron analysis reveals that between 2023 and 2025 achievement raise investment rounds for a total of 37.5 billion dollars (20,000 of them in 2025 alone). A good part of that money came precisely from the companies that provide infrastructure: Amazon and Google. Despite that funding, Anthropic appears as desperate as OpenAI to raise new rounds of investment. The company run by Dario Amodei recently resorted to money from Middle Eastern countries, for example. Spending continues to skyrocket. The study figures further reveal that Anthropic spends more the more time passes. In January 2024, it spent $52.9 million on AWS, but in December 2024 that amount rose to $176.1 million. In September 2025, it is estimated that spending on AWS was no less than $518.9 million: the escalation in costs is very notable. And he tightens the screws on Cursor. One of Anthropic’s most important clients is the startup vibe coding Cursor. This company has clearly been affected by that situation, and Cursor’s costs on AWS doubled from $6.19 million in May 2025 to $12.67 million in June. Just in those Anthropic months implement the so-called “Service Levels” with which it forced business customers to spend a minimum amount and pay higher rates for prompt caching, a special component designed for startups that use generative AI models for programming. What did Cursor do? Increase prices (and apologize for it) of your customer subscriptions. This can’t go on like this forever. For Zitron, always very critical of this reality of AI companies, the conclusion is clear: Anthropic’s costs are out of control. In fact, he argues that they increase practically linearly with respect to revenue, which makes their business model unsustainable. The only solution is to increase prices drastically (possibly 100%) to become profitable. The problem is that the market accepts paying twice as much at once for AI as it currently pays for. Image | Anthropic | Taylor Vick In Xataka | Anthropic says Claude Sonnet 4.5 can clone a service like Slack in 30 hours. The reality is more complicated

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.