claims

Anthropic claims to have found Claude’s “consciousness”, although the data tells another story

by

Anthropic has published an article that promises to change the way we understand the inner workings of your business models. artificial intelligence (AI), like we explain to you last week. This company claims to have identified a kind of global workspace within Claudewhich he has named J-Spacewhere the model analyzes and manipulates concepts before these become part of the final answer. What is surprising is that, according to Anthropic, this space has not been deliberately programmed: it has emerged as a byproduct of training the model. To analyze it and understand how it works, a technique known as J-Lens which is capable of mapping the internal activations of the model onto words from its own output vocabulary. When asking Claude to solve a multi-step mathematical calculation the visible answer contained only the final result, but within the J-Space each intermediate step was represented separately. Without this technique no part of that process could have been observed from the outside. However, Anthropic itself recognizes that much of what a language model does does not even go through this internal space. Speaking fluently, remembering simple facts or applying basic grammatical rules almost completely dispenses with J-Spaceand when the company blocked its use Claude continued to function normally, although he lost some of his higher-order cognitive functions. Claude “knows” that he is being watched During one of his most striking experiments, Anthropic subjected Claude to injection attempts. prompts during data acquisition processes. He J-Space then showed words like “false”, “injection”, “erroneous”, “prompt“, “fraud” and “poison”, while the visible response of the model completely ignored the attempted manipulation. Interestingly, when pressed to comment on the matter, terms such as “transparency”, “honesty” and “truth” emerged in the internal space. The language used in the article sometimes borders on the idea of ​​an emerging consciousness The Anthropic researchers also observed that Claude seems to recognize when he is being evaluated, which leads him to behave differently when faced with tests than when faced with prompts everyday. In a particularly revealing test, Anthropic removed J-Space the vocabulary associated with that consciousness of being put to the test (terms such as “fake”, “fictitious” or “scenario”) and found that Claude became much more vulnerable to attempts at blackmail or provocation. Be that as it may, the researchers’ report admits important limits: the J-Space is restricted to a single-token vocabulary, so plans or concepts that cannot be expressed with a single word may not appear in the reading, even though the model is calculating them anyway. Anthropic also does not dare to claim that monitoring this space is sufficient to guarantee the alignment of a model. However, we must not overlook that the language used in the article sometimes borders on the idea of ​​an emerging consciousness. Neel Nanda, head of model interpretability at DeepMind, supports this finding as real evidence of a cognitive space within the modelsalthough he clarifies that the practical usefulness of J-Lens remains limited. This achievement, in any case, opens a promising avenue for auditing the honesty of models, although it is still far from being a complete window into a machine’s thinking. Image | Generated by Xataka with ChatGPT using a prompt created with Claude More information | Anthropic In Xataka | While most oppose AI data centers, there is one group enthusiastic about them: merchandise thieves

We thought the MacBook Neo was the perfect affordable computer. AMD claims it has a big problem: games

by

He MacBook Neo has been everything a missile on the waterline of the “affordable” Windows laptop market. Our analysis confirmed our expectations and we were able to verify that we are facing a more than solvent team with a outstanding price/performance ratio. However, it is not perfect, and its Achilles heel is in a specific area: video games. AMD strikes back. The company has launched an aggressive advertising campaign with a strong message: “The competition makes sacrifices. You don’t have to make them”, and then focus on how the Apple team falters in a section that is important for a notable sector of users: “While 15 of the top 20 PC games don’t run natively on the MacBook Neo, AMD systems give you access to a massive library of games across Steam, Epic Games Store, and PC Game Pass.” The hateful comparisons. In that campaign AMD compares the HP Omnibook X Flip with an AMD Ryzen 5 220 with Apple’s MacBook Neo. It boasts of its 512 GB SSD (256 GB in the Apple model indicated, although there is a 512 GB version), its touch screen and its greater connectivity options. Beware of AMD’s message. As noted in Tom’s Hardware, this AMD chip is not new, but rather a version derived from the 8540U that is accompanied by an integrated Radeon 740M GPU. It’s a modest chip that can run titles like GTA V at 100 FPS in low quality, but Hellblade 2 runs at 8 FPS and Alan Wake 2 runs at 11 FPS. So it may indeed be able to run those video games, but it’s not clear that they are actually “playable.” One thing is certain: the MacBook Neo is not for gaming. In reality, Apple does not focus its MacBook Neo on gamers, because it knows that although the A18 Pro is a truly remarkable SoCis not designed for that sector. The company has its Metal API to be able to run games and it is even possible to enjoy some native ones with some joy, but if what you want is to play (especially with top PC titles), the MacBook Neo is not your device. The laptop war for students. This counterattack from AMD is to be expected: the company has undoubtedly been affected by the enormous success of the MacBook Neo, and is trying to demonstrate that its proposals are just as valid or even better. The juicy niche of student laptops is at play here: if you convince a young person to opt for a MacBook Neo, they will likely end up trapped in the Apple ecosystem (if they weren’t already). Intel also reacts. Intel’s response to the MacBook Neo has not been an advertising campaign, but something much more tangible. The company has announced its platform Wildcat Lakechips made with 18A photolithography that are a priori up to 21% more powerful than the A18 Pro. Equipment like the Chuwi Unibook (from $449) are better at least with the specifications sheet in hand, but its real performance is for now an unknown and we will have to wait for independent analyzes. More options are coming. The Computex fair that took place two weeks ago also made it clear that other manufacturers they are going to try to take advantage this moment of uncertainty to sneak in their proposals. Qualcomm launched its Snapdragon C and Nvidia made something similar with their RTX Spark (we will see the prices). What is clear is that the launch of the MacBook Neo has awakened a segment that had become comfortable, and that is good: there will be a lot of competition and, above all, many options. In Xataka | Apple is selling so many MacBook Neos that it runs the risk of not being able to make more

A US company claims it can build a cutting-edge lithography machine. ASML says not even remotely

by

Substrate is not just another startup. It was founded in 2022 by brothers James and Oliver Proud, and is backed by Peter Thiel (he co-founded PayPal and is one of the largest investors in Silicon Valley). Despite having existed for only four years, it has raised more than 100 million dollars and has been valued at more than 1 billion. This very successful start-up is based on a promise: the Proud brothers claim that they can build photolithography equipment as advanced as the most sophisticated they have. the Dutch company ASML. Currently this firm from the Netherlands is the only one capable of manufacturing the machines extreme ultraviolet lithography (EUV) that are used to produce cutting-edge semiconductors, which has placed it in an effective monopoly position in the global semiconductor industry. In the current scenario of confrontation with China, the US is interested in having a national company capable of manufacturing cutting-edge lithography equipment. This is Substrate’s trump card. However, we have reasonable grounds for reluctantly taking up the promise of the Proud brothers. “No one is coming for us” Christophe Fouquet, the general director of ASML, assures that no company on the planet is in a position to compete with them. During a conversation with Connie Loizos, an editor at TechCrunch, Fouquet has argued that “the challenges of lithography are many. Being able to make an image is a starting point, but that image must be produced in large quantities, at very low cost, at high speed and with nanometric precision.” It makes sense. “We had to solve only one problem: obtaining extreme ultraviolet light. And that alone took us 20 years” “I always say that the only reason ASML was able to build an EUV machine is because 80% of it already existed from prior knowledge and products developed over time. We had to solve just one problem: getting the extreme ultraviolet light. And that alone took us 20 years. When starting from scratch, the challenge is enormous. I have heard many statements. And I have seen some images. But we got our first image with EUV technology 30 years ago, and even then we need 20 more years of hard work to turn it into a manufacturing system,” points out the head of ASML. It is clear that Christophe Fouquet trusts his technology. And in your product. However, the starting point of Substrate is different from that of ASML. This American startup uses a particle accelerator as a light source for an X-ray lithography tool instead of using extreme ultraviolet light like ASML. According to the Proud brothers Their technology allows them to manufacture a silicon wafer at an order of magnitude lower cost than with ASML’s EUV approach. Be that as it may, there is another fundamental difference between the ASML and Substrate strategies. And instead of supplying machines to chip manufacturers, as ASML does, Substrate wants to establish its own network of semiconductor production plants equipped with its photolithography machines. Furthermore, its plan is very ambitious: it aims to produce cutting-edge integrated circuits on a large scale in 2028. It sounds daring, no doubt, but time, as always, will put everything in its place. Image | ASML More information | TechCrunch In Xataka | TSMC has made the chip industry’s most intriguing decision: not to use ASML’s most advanced machines

The NYT claims to have found Satoshi Nakamoto and the evidence is as conclusive as ever. That is, little or nothing

by

On October 31, 2008, someone calling himself Satoshi Nakamoto posted a document nine-page document that proposed a decentralized electronic money system. That was the birth of bitcoin and a phenomenon that has shaken the foundations of the modern economy, but no one has ever been able to know who was really behind that pseudonym. And no matter what they say in The New York Times, we still don’t know. Adam Back. The journalist John Carreyrou became famous for uncover he Theranos fraudand in a new and in-depth investigation published in The New York Times claims to have discovered the real identity of Satoshi Nakamoto. According to their data, that person is Adam Back, a well-known 55-year-old British cryptographer who is currently the CEO of the Blockstream company. Back denies all these claims, and we fear that this investigation will once again end up like the rest of the previous attempts. He is not just any suspect. In 1997, Back invented HashCash, a mathematical puzzle-solving system that Satoshi explicitly cited in his white paper and that in fact became the basis of the bitcoin mining mechanism. He was an active member of groups of anarchist cryptographers who had been trying to create digital money for decades. And between 1997 and 1999 he published messages on Internet forums that accurately described almost all the fundamental elements of the system. That was like an instruction manual for creating bitcoin, but created a decade before Satoshi Nakamoto published it. Analysis based…on how Satoshi wrote? The methodology followed by Carreyrou is striking. He worked with an AI expert from the NYT to collect messages from crypto mailing lists between 1992 and 2008, merged them into a database with 134,000 messages from 620 different users, and applied three methods of analyzing those messages. In the analysis, it was seen that the British spelling, the same hyphen errors in some terms, the confusion between “it” and “its” or the use of two spaces between sentences pointed to Back writing the same as Satoshi in his document. All circumstantial. This style analysis led Back to be identified as the clear candidate among a group of 12 “suspects.” However, the experts who carried out the analysis were not entirely sure and, for example, another of the traditional candidates in these investigations, Hal Finney, was almost tied in these style matches. Back himself indicated that anyone who writes about cryptography is going to sound just like the rest of the community. Back denies the major. Carreyrou ended up traveling to a bitcoin conference in El Salvador to meet with Back and explain his findings. During their two-hour conversation, Back denied being Satoshi on several occasions. The journalist claimed to have caught Back making a mistake when he stated that “I’m better with code than words”, something that Satoshi Nakamoto also said in a late 2008 message on one of the mailing lists analyzed. Back said he was simply making a general observation about programmers, and the truth is that all of Carreyrou’s evidence was inconclusive. Only Satoshi can prove that he is Satoshi. The bitcoin user community often says that “we are all Satoshi”, in reference to the fact that revealing the real identity of the creator of bitcoin is not relevant. The only way to do this would be for Satoshi himself to reveal himself and prove to be that person by transferring bitcoins from one of Satoshi Nakamoto’s original wallets – where he houses nothing less than 1.1 million bitcoins. Many have tried to unmask Satoshi. They have all failed. We are facing another attempt to reveal the real identity of Satoshi Nakamoto. It is neither the first nor probably the last. In recent years we have seen several more: Neither then nor now has anything really been proven. The mystery continues. Image | Wikipedia | Michael Fortsch In Xataka | The creator of Bitcoin disappeared 13 years ago. Now your account has just received a million-dollar transfer

China claims that having its own ASML is not that difficult. At the end of the day “ASML is a simple integrator”

by

China is advancing at a fearsome speed, but they are their own critics. A group of scientists and industry representatives has published an analysis very critical in which they describe their semiconductor industry as “small, dispersed and weak.” For these experts, the problem is that they do not have their own ASML, so they propose creating one, and the curious thing about it is precisely that: that they do not see it as difficult to do. Not authors, authorities. Among the authors of the study are Wang Yangyuan —co-founder of SMIC— or Chen Nanxiang —director of the YMTC NAND chip manufacturer—. Both they and the rest of the participants are recognized personalities in the field of semiconductors. That makes it clear that this articlepublished in the magazine Science and Technology Review China is valuable to understand the state of this industry. The US veto works. This article indicates that the United States has managed to contain China’s advance in three areas: electronic design automation (EDA) used in chip design, the manufacture of silicon wafers, and the creation of chip manufacturing equipment, especially those with extreme ultraviolet (UVE) photolithographic technologya segment that is absolutely dominated by the Dutch ASML and that China has not yet managed to match despite its efforts. There are many companies, but they are too small. The study analyzes in depth the situation of the Chinese semiconductor market, which they describe as small, dispersed and weak. And to prove it they give a significant fact: there are 3,626 domestic chip design companies in China, but “the total value of the industry’s output was 646 billion yuan (approximately 91 billion dollars). In other words: the total sales of these 3,626 companies were less than the sales of NVIDIA alone.” Smartphone chips are doing well. Of course, these experts point out that several Chinese companies have managed to reach “the world’s leading edge” in terms of chips for smartphones. They mention HiSilicon Semiconductor and Unisoc, which occupy the first and second place among the largest smartphone chip designers in China, with market shares of 20 and 10% respectively. And the “mature” chips, too. Something similar happens with chips with much more mature technologies, such as those manufactured with 28 nm photolithography or higher. That problem is already solved in China, which does not have to depend so much on foreign manufacturers. In fact, China now represents 33% of global production in this segment, and design and manufacturing processes are not limited by internal restrictions. Too dependent. Despite its many advances, China remains the world’s largest importer of integrated circuits. It invested $385.79 billion in 2024 in these components, even exceeding its oil imports ($324.7 billion). Here China has a big problem of dependence on chips from third-party foreign manufacturers, and this is especially noticeable in automotive chips (95% of them import) and memory chips (90%): all of them depend on imports. There is a clear bottleneck in high-end integrated circuit production. In search of Chinese ASML. The document also addresses competition with ASML but not as a direct commercial rivalry, but as a strategic challenge of technological sovereignty. The Dutch company is described as “a simple integrator” that coordinates more than 5,000 suppliers to manage the 10,000 components of an EUV machine. The suggestion of the study is precisely to create a Chinese ASML that unifies the advances made by different companies. But. Although progress has been made In this ambitious objective of creating SVU machines, “integrating them with a national effort is a problem that must be resolved during the 15th five-year plan” that concludes in 2030. To do this, these experts assure, there must be financial support and human personnel is required. We already knew that China I was trying to copy ASMLbut for now it is going badly. The document talks about developments such as promising Flip-FET technology (FFET) of Peking University. This advance allows us to reach 3-2 nm nodes without depending exclusively on EUV machines, but it remains to be seen if this method ends up being successful or not. In Xataka | Holland has just declared war on China in the most important battle of the century: control of semiconductors

its electrical grid claims to be “full” when in reality it is underutilized

by

Spain is experiencing an obvious and costly energy paradox. While the country breaks renewable generation recordsits electrical system suffers an administrative “thrombosis” that threatens to stop reindustrialization. The problem is that the system works like a broken bridge: clean energy is born in the so-called “emptied Spain”, but there are not enough cables to take it to the cities and factories where consumption is concentrated. The panic in the sector reached its peak when the National Markets and Competition Commission (CNMC) was forced to postpone three months (from February 2 to May 4, 2026) the publication of the access capacity maps after a critical alert from Red Eléctrica: under the new security criteria, approximately 90% of the network nodes would appear in “red”, that is, with zero capacity. However, the network is not physically collapsed, but administratively “full” and underutilized in practice. To solve this funnel, the CNMC has put on the table a master plan that will change the rules of the game: flexible access permissions. The perfect storm. Getting to this point has not been the result of a single mistake, but rather a cocktail of bureaucratic slowness, territorial imbalances and speculation. As we have already advanced in Xataka, There is a huge gap between administrative times and physical execution: building a substation barely requires a year of work, but its prior processing can take between three and six years. Added to this is that we have installed windmills and solar panels where there is land and resources, but demand is growing in metropolitan areas that do not have sufficient infrastructure, leaving 83.4% of distribution nodes saturated currents. The consequences on the street are devastating. Last year only 12% of connection requests for new urban developments were granted, which, according to the Asprima employers’ associationputs the construction of 350,000 homes at risk due to the simple lack of electrical power. And in the midst of the chaos, the bubble: there are access requests for 67,100 MW (half of all the installed power in the country), which makes the regulator suspect the existence of “ghost” projects that hoard nodes only to resell the permits. The end of the binary model. Until now, the electrical system operated under a binary principle: either they gave you firm access, 100% guaranteed, or they denied it. However, as he noted on his social networks the Secretary of State for Energy, Joan Groizard, the current network is underused; In fact, a “smaller” network in the past supported demand peaks much higher than today. This is where the regulatory revolution comes in. The CNMC proposal breaks with the resounding “no” and establishes that, if there is residual capacity at certain times of the day or year, it can be shared. Flexible access capability assumes that supply will not be guaranteed at all hours of the year, maximizing the use of existing infrastructure without immediately resorting to massive investments that citizens would end up paying for. The four ways of flexibility. To articulate this new paradigm, the supporting report and the proposed resolution of the CNMC define four types of permits Flexible access, adapted to different needs: Permission Type 0 (Fixed pattern in Distribution): Applies to installations connected to any voltage level in the distribution network. It allows energy to be consumed following a fixed time pattern (for example, from 00:00 to 07:59 and from 11:00 to 17:59), which represents at least 62.5% of the hours of the year. Outside of these ranges, if the installation consumes power, the network manager (GRD) can disconnect it remotely without prior notice. It is ideal for those who can plan their production. Type 1 Permit (Remote disconnection due to contingency N-1): Designed for distribution installations with voltage greater than 36 kV. The installation meets the requirements under normal conditions (with an expected consumption of 90% of the year), but agrees to be disconnected remotely and without prior notice if any element fails in the substation itself to which it is connected. Type 2 Permit (Dynamic Instructions in Distribution): For voltages greater than 36 kV and powers greater than 1 MW. It is the most technologically advanced, the installation must be able to receive dynamic instructions from the GRD to reduce its load, whether scheduled the day before or in real time. Response times are critical: less than 30 minutes if it is preventive, or less than 3 minutes (immediate) if it is corrective. If you disobey, you will be disconnected. This model will come into force from January 1, 2028. Type 3 Permit (Transmission with automatic reduction): Aimed at demand installations of more than 1 MW connected directly to the transmission network. These facilities are required to participate in the Automatic Power Reduction System (SRAP). In the event of a security alert from the electrical system, the operator (OS) will send a signal and the installation must reduce the power associated with its flexible access to zero effectively and immediately. Winners, exceptions and the bill. This regulatory change has clear winners and some red lines. Storage facilities in demand mode are the perfect candidates, since, by law, they have no guarantee of supply and will have 7 months to request the modification of their permits. At the opposite extreme, the CNMC explicitly prohibits granting these accesses to essential supplies (such as hospitals), to demands that do not support 24 hours without a network, and to collective projects such as urban plans. Modernizing the network to support this “reinforced mode” of digitalized operation will have a direct economic impact. The forecasts for 2026 point to increases in citizen receipts of 4% in tolls and 10.5% in charges to finance system adjustments. For now, the clock is ticking: the public hearing process for agents to send their allegations to the CNMC proposal will end on March 20, 2026. Connect the future. Spain finds itself at an ironic and dangerous crossroads. The country has everything to be the great green battery of Europe, but the lack of cables and excess bureaucracy … Read more

claims its new humanoid robot runs like an elite athlete

by

For years, when we think about humanoid robots, the image that usually comes to mind is that of machines capable of making impossible jumps or executing perfectly measured stuntslike those popularized by Boston Dynamics with Atlas. At the opposite extreme are designs oriented toward repetitive tasks and controlled environments, like Tesla’s Optimus. Between both imaginaries there is an obvious gap: it is difficult to visualize a robot as an athlete. However, that frontier is beginning to move with proposals that no longer only seek balance or skill, but also performance typical of human athletics. ‘Bolt’, the athlete robot. We are talking about the new humanoid presented by the Chinese company MirrorMe Technology, which claims to have reached a maximum speed of 10 meters per second during tests in real conditions. The company maintains that this is the first time that a full-size robot has achieved that record outside of a laboratory, a milestone that, if confirmed, would shift the conversation from controlled demonstration to performance in environments closer to the physical world. To place the magnitude of the data as a historical reference, Usain Bolt clocked 9.58 seconds in the 100 meters during the 2009 World Cup in Berlin, a mark that at that time defined the ceiling of sprint athletics. Real running. Achieving high speeds in a humanoid requires solving one of the biggest challenges in bipedal robotics: maintaining dynamic balance while the body endures repeated impacts and constant changes of support. In humans, this coordination between perception, motor control and response occurs almost automatically, but in a machine it requires redesigning joints, optimizing energy delivery and fine-tuning stability in real time. MirrorMe maintains that the Bolt incorporates new joint configurations and a power system optimized to approximate human movement patterns. A job that started years ago. The company had already attracted attention with Black Panther IIa research-oriented robot that traveled 100 meters in 13.17 seconds during a television broadcast in China. In that same demonstration its top speed was around 9.7 meters per second. Bolt thus appears as the next logical step in that search for ever-increasing physical performance. Robots as athletes. Bolt’s advance also fits into a broader context in which China is exploring the athletic dimension of humanoid robotics. Public demonstrations of robots have appeared in recent years capable of fighting in disciplines such as kickboxingdeveloped by companies such as Unitree Robotics, in addition to viral tournaments that serve as a showcase to measure agility, balance and coordination. This ecosystem suggests that physical performance is beginning to become a relevant metric beyond pure research. The future. Beyond the record that the company says it has achieved, MirrorMe imagines specific applications for this type of high-performance humanoids. Among them, he mentions the possibility of them acting as training partners for human athletes, an idea that points both to professional sports and to research in biomechanics and movement. However, as with many announcements in advanced robotics, the true scope will depend on sustained testing over time and real-world usage scenarios. Images | MirrorMe In Xataka | OpenAI going from 70% share to 46% is the symptom of something more worrying: they have entered panic mode

A Chinese startup claims to have created its own TPU to compete with NVIDIA. The only problem is that it is three years late

by

A Chinese startup called Zhonghao Xinying (known internationally as CL Tech) has come to the fore with a bold promise. The company claims to have developed an AI chip that not only circumvents Western intellectual property restrictions, but also outperforms NVIDIA’s A100 chip. Which is very good, but also a little bad. Chana arrives. The chip in question has been named “Chana”, and according to SCMP we are dealing with a GPTPU (General Purpose Tensor Processing Unit). Unlike NVIDIA GPUs, aimed at accelerating AI workloads, this is an ASIC, that is, an application-specific integrated circuit designed from the ground up for neural network workloads. promise. According to Zhonghao Xinying Chana, it offers up to 1.5 times the performance of the NVIDIA A100 based on the Ampere architecture. Not only that: it achieves that performance with 30% lower consumption. The startup highlights that the computational cost per unit would therefore be less than half of that offered by the A100 chips. A little history of the company. Behind Zhonghao Xinying is Yanggong Yifan, an engineer formed at Stanford and the University of Michigan. He worked on the development of several generations of Google TPUs and also on the development of Oracle chips, and in 2018 founded this startup in Hangzhou together with Hanxun Zhengan engineer who worked at Samsung for several years. They were joined by other engineers from Microsoft, Oracle, NVIDIA, Amazon and Facebook, they indicate. on Baidu. We are therefore faced with several of those cases of “boomerang talent” with Chinese engineers who are forged in the US and then return to China to create solutions for their own industry. Solutions that do not depend on the West. Yanggong affirms that its chip features “fully self-controlled IP cores, a custom instruction set, and a fully in-house computing platform. Our chips do not rely on foreign technology licenses, ensuring long-term security and sustainability from an architectural perspective.” But. Although the achievement is striking, it is necessary to put it in perspective. The NVIDIA A100 is a 2020 AI GPU, and even with the improvements that this Chinese startup promises, its performance is, for example, far from H100 chips with Hopper architecture that appeared in 2022. Not to mention of the latest Blackwell Ultra chipswhich are currently NVIDIA’s greatest exponent in terms of AI chips. There are also no details about who makes the chip, and one of the candidates it would be SMICwhich has 7nm technology. They are very far away, and they have another problem. The technical achievement of these engineers is certainly notable, but everything indicates that they are still far from what NVIDIA and its competitors are achieving. like AMD or Google with its recent TPU Ironwood. There is another element that works against them: Chinese manufacturers continue without having direct access to the most advanced photolithography on the market, and although it also there is progress from Chinese manufacturers in that sense, competing is certainly complicated without access to the most advanced technologies. Pressure. In 2024 the company achievement revenues of 598 million yuan (73 million euros) with a net profit of 85.9 million yuan, but in the first half of the year the income was only 102 million yuan and had losses of 144 million yuan. The firm has reached an agreement with its investors by which it will have to go public at the end of 2026, or else it will be forced to buy back shares. The financial pressure is therefore notable for the company, which must demonstrate in the coming months that its roadmap is truly competitive. In Xataka | China was no longer supposed to be able to get its hands on NVIDIA’s most advanced chips. Until he found a shortcut in Indonesia

Grokipedia claims to aspire to the truth. An investigation has just shown that he cites neo-Nazi forums and conspiracy websites

by

The proposal of Grokipedia came accompanied by a direct message: aspire to “the truth and nothing but the truth,” as stated by Elon Musk in X. That statement takes on a new context. after the publication of a Cornell Tech study which examines how various entries are constructed and what fonts they use. The analysis shows that, along with content inherited almost literally from Wikipedia, there are articles that use sources cataloged by academic institutions and verification organizations such as neo-nazi spaces or openly conspiratorial sites. At first glance, Grokipedia takes on a familiar appearance: a home page dominated by a search engine and articles with headings and references. The inner workings, however, are much less transparent. Users do not have a clear system to suggest changes and, at the top of some entries, the label “Reviewed by Grok X weeks ago” appears, indicating an intervention by the AI ​​chatbot without detailing criteria or those responsible. In Wikipediathe edition history is public and allows each modification to be reconstructed. Grokipedia under the magnifying glass The aforementioned analysis compares both platforms on a large scale and points out that, although Grokipedia publishes longer articles and with twice as many citations as Wikipediamuch of its content comes from there. Of course, the coincidence varies: pages with a Creative Commons (CC) license present very high similarities, while those generated without that license are further removed from the original. One of the most delicate issues is the appearance of references to controversial platforms. InfoWars, which according to the authors is not cited even once on Wikipedia, has 34 mentions on Grokipedia. The pattern is repeated with other low credibility domains: Stormfront reaches 42 citations, LifeSiteNews reaches 100 and the Global Research and VoltaireNet sites register 51 and 45 references respectively. All of them are practically non-existent on Wikipedia, reflecting clear differences in source selection filters. Elon Musk’s entry in Grokipedia To mention a few examples, Leiden University characterizes Stormfront as a forum associated with right-wing extremism already current neo-naziswith a founder linked to Ku Klux Klan and a trajectory mentioned in several studies for its relationship with violent incidents. PolitiFact, on the other hand, defines Infowars as a portal dedicated to conspiracy theories and run by Alex Jonesa presenter known for promoting this type of content. This is what the edition history looks like in Grokipedia What appears in the study is not limited to counting how many times these domains are cited. It also highlights that the presence of sources considered unreliable or directly discarded by Wikipedia is much more widespread in Grokipedia. And one of the authors, in a text published in Indicatorcollects this accumulation of low-quality references to describe a broader pattern: Grokipedia seems to be making its own editorial decisions that alter the focus of certain topics. It remains to be seen how Grokipedia will evolve and what publishing model it will adopt as it grows. No encyclopedia works as a perfect reference —neither Wikipedia nor Grokipedia—, but they do operate with different mechanics. As we say, Wikipedia relies on an open community with standards, public debates and an accessible history of changes; Grokipedia, on the other hand, is based on criteria that are more difficult to follow from the outside, with an AI assistant that intervenes in the texts and without a clear human collaboration system. Images | Gage Skidmore (C BY-SA 4.0) | In Xataka | Carnegie built libraries, Gates sold them on CD-ROM, Musk locked them in an AI: the history of knowledge control

A company claims to have created the first transparent computer monitor: it has borrowed technology from aviation

by

The idea of ​​“seeing through” the screen is not new. For decades, HUDs have projected data onto the windshields of airplanes. and today many cars offer own versions to show speed or navigation without taking your eyes off the road. That principle, project or reflect information on a transparent elementis the one that Visual Instruments ensures have brought to the desktop with Phantom, a monitor that seeks to mix the digital and the physical and that adjusts its transparency in real time. Phantom does not use a panel that becomes transparent on its own, but rather an optical assembly similar to that of a teleprompter or HUD. The image is reflected in a tilted glass and thus appears to “float” over what is behind, with an opacity control that allows it to go from transparent to opaque. The company presents it as “the first transparent computer monitor.” The key to the invention is in the optics, not in a futuristic panel According to the manufacturer, Phantom is presented as a 24-inch monitor in 16:9 format with 4K resolution. The company sets the brightness in a range from 5 to 5,000 nits and places the color coverage at 100% sRGB, figures that must be confirmed when the product reaches the hands of users and analysts. As for the connection, it is limited to the most common options: USB-C with DisplayPort and HDMIwithout the need for additional software to operate. {“videoId”:”x87bool”,”autoplay”:false,”title”:”How to CALIBRATE your MONITOR for VIDEO GAMES and enjoy it to the fullest”, “tag”:”monitor”, “duration”:”230″} It is worth putting it in context. Transparent screens have been appearing in consumer markets for years. LG has been marketing a 77-inch transparent OLED TV since 2024. Lenovo, for its part, has also developed a concept laptop with a transparent panel. The difference here is key: those products use transparent base panels, while Phantom uses, as we say, HUD-type optics. They are different technological routes, with different commitments and limitations. Aspects such as contrast, color representation with different ambient lights, the presence of reflectionsuniformity or comfort in prolonged use remain unknown. It is also unclear to what extent alternating between the monitor and the background can translate into real relief from eye strain beyond the idea put forward by the manufacturer. Everything will depend on the first independent tests and the measurements that are published when demonstration units exist. In Xataka There is a cheap TV sweeping Amazon: after a week of use it became clear in which situations it can make sense Phantom arrives, therefore, as a suggestive idea that has yet to demonstrate whether it can be sustained in everyday use. The commercial status points to a very early and selective launch. The Founders Edition is limited to ten units intended for early adopters, with shipping expected in Q4 2025, 30-day returns, and a one-year warranty. The price is not final: each unit is custom configured, but the company itself compares it to an Apple Studio Display whose price in Spain starts at 1,779 euros. Images | Visual Instruments | Telstar Logistics In Xataka | Xiaomi TV S Pro MiniLED 2026, analysis: the juggling act of wanting to offer the quality of a high-end and the price of a mid-range (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 A company claims to have created the first transparent computer monitor: it has borrowed technology from aviation was originally published in Xataka by Javier Marquez .

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.