The new Chinese model Kimi K3 is already number one in Frontend Code Arena. And it’s unleashing madness on the Internet

It seems like yesterday when DeepSeek R1 called into question an idea that many took for granted: that the race for advanced artificial intelligence It still had a clear owner in Silicon Valley. The emergence of the Chinese model helped trigger a massive sale of technology and led NVIDIA to suffer a loss daily capitalization unprecedented until then. As the months passed, that image lost intensity, but the message remained: the Chinese technological ecosystem was not willing to limit itself to keeping pace with the United States. The following notice now has a different name: Kimi K3. Moonshot AI has just presented a model with 2.8 trillion total parameters that, as soon as it arrived, was placed at the top of Frontend Code Arenaahead of some of the most powerful proposals from Anthropic and OpenAI. But the story is not limited to a classification: developers and fans are already using it to create interfaces, games and recreations that anyone can see and, in some cases, try. That’s where this article really begins. It is worth dwelling on the details of that classification. At the time of writing, Kimi K3 reaches 1,679 points in Frontend Code Arena, ahead of Claude Fable 5with 1,631, and GPT-5.6 Sol xHigh, with 1,618. The improvement compared to the previous generation is also striking: Kimi K2.6 was in 18th placewhile his successor leads six of the seven domains evaluated. For now, Arena maintains the label of preliminary result, so it is convenient to read this position as a very significant photograph, but still susceptible to change. We are not facing a universal programming exam, but rather a very specific test. Frontend Code Arena compares web applications created by different models and lets users evaluate which one solves the task better, which one works more reliably, and which one presents a better experience. That approach is especially useful for measuring visible and practical capabilities, but it also has obvious limits. That Kimi K3 leads here tells us a lot about its frontend performance, although it doesn’t automatically allow us to extend that advantage to complex repositories, backend, mathematics, or general reasoning. Outside of this specific terrain, photography remains favorable, although more balanced. Vals AI places Kimi K3 second among 38 models, with 74.70%just behind Claude Fable 5, which reaches 75.14%, and above GPT-5.6 Sol, with 73.12%. Artificial Analysis also places it among the most advanced systems in its classification, with 57 points and third place overall. Where Kimi K3 seems to feel most comfortable is in tasks that combine programming, visual context and several chained steps. Arena supports its ability to build web interfaces, while Vals AI also records high performance in agent programming tests. Moonshot adds that the model can traverse large repositories, use terminal tools, and review screenshots of its own work to correct the output on the fly. That last capability, which the company calls “vision in the loop,” helps explain why it excels at transforming visual references into interactive products. There are also several cautions before interpreting Kimi K3 as a definitive victory. Moonshot presents it as an open weight model, but those files have not been published yet and the company promises to release them no later than July 27. Nor should we confuse this openness with complete open source, because details about the license and the rest of the system are still missing. Its 2.8 billion total parameters belong to a sparse architecture that activates 16 of its 896 experts. The company itself recommends configurations with 64 accelerators or more, very far from what a conventional computer can offer. The community reaction helps understand why Kimi K3 is attracting so much attention. One of the most striking examples is a recreation of macOS 27 which works within the browser and which its creator attributes to a swarm of model agents working for about three hours. They add to it Ballista, an interactive panel with a 3D balloon and several comparisons against Claude and GPT. They are not independent benchmarks, but demos shared by their own creators, but they allow you to see what kind of results the model is producing outside the tables. To create something like the macOS simulation or the ballista game, we don’t need to model every element by hand from scratch. We can describe the resultattach a reference and commission Kimi to build a functional application, for example with HTML, JavaScript and various graphics libraries. The project is then tested, modified, and finally published or recorded for sharing. Kimi K3 can be used from Kimi.comKimi Work, Kimi Code or tools connected to its API, although it is not confirmed which specific environment was used in several of the examples we have seen. It is still early to turn this launch into a definitive change of leadership. Fable 5 and GPT-5.6 Sun They are still ahead in several evaluations, the Kimi K3’s weights are not yet available and many of its capabilities will have to be verified with more time. Even so, what we have seen is already difficult to ignore: a Chinese company can compete for leading positions, offer competitive results and get the community to transform that capacity into real applications almost immediately. The race continues, but the margin between its main protagonists seems increasingly narrower. Images | Kimi | Screenshot In Xataka | China has a plan to win the AI ​​war against the US. And DeepSeek is its champion

The visual garbage of AI is so omnipresent that it is already unleashing a counter-aesthetic current: neo-brutalism

The Internet is being flooded with images and designs that seem to be cut from the same mold: identical fonts, predictable gradients, aesthetics polished to the point of nausea. This phenomenon is difficult to describe and limit due to its infinite variants and omnipresence, but it has a name: “AI slop“. By this we refer to digital content generated with artificial intelligence, from images to web design itself, and where quantity takes precedence over any hint of originality or meaning beyond the effectiveness of the mass production chain. But what is AI Slop. The expression gained traction in 2024 thanks to British programmer Simon Willisonalthough it had previously circulated in communities such as 4chan and Hacker News. The concept indicates a root problem: When AI models are trained with the most common patterns on the internet, they replicate a generic and forgettable aesthetic ad nauseam. It’s what experts call “distributional convergence”: everything seems designed by the same depersonalized algorithm. And the anti-AI slop? Faced with this invasion of algorithmic uniformity, a visual counterculture emerges that celebrates precisely what AI avoids: the clumsiness, the unevenness, the marks of the human creative process. The anti-AI slop is not an aesthetic whim, but a declaration of principles that rescues imperfection and turns it into a differential value and a trait of delicious humanity. Some critics celebrate it as a kind of digital neo-brutalism, referring to the famous unadorned concrete architecture of the 1950s. This neo-brutalism is characterized by taking digital nudity to the extreme: sites built with basic HTML and minimal CSS, where the code is displayed without artifice. The fonts are not the elegant paid fonts, but the system ones installed by default: Arial, Times New Roman, Courier. The photographs appear unretouched, with their digital noises and compression artifacts clearly visible. Asymmetrical compositions, in short, that break any notion of classical balance. Like children. This leads us to a style perhaps opposite to cold brutalism, but also contrary to IA Slop: the aesthetic of a childish hasty sketch. Deliberately unbalanced proportions, freehand illustrations, elements that overflow the margins. Lindsay Marsh, a designer specializing in visual trends, points out that These visible “errors” act as signatures of authenticity: They are proof that behind the screen there are human fingers, not processors without humanity. The people of Phantom Watchers formulates it in a similar way: “It’s our way of saying ‘a human was here.’” Any notable example? The recent redesign of the oldest magazine The Face It is full of imperfections. Hell, it even looks like they programmed it in HTML. What features does it have? Like IA Slop itself, this opposition mutates in countless ways: disproportionately large fonts that challenge traditional visual hierarchy, website scaffolding exposed in an exhibitionist manner (even leaving the code visible), and color combinations limited to one or two colors on uniform black or white backgrounds, sometimes imitating the texture of analog montage. The templates are twisted on purpose, breaking with the obsessive symmetry that dominates more formal styles, and which are easier to imitate by those AIs that propose to set up a web store in just a few minutes and with a couple of prompts. But… why? The guiding principles of this rejection movement are clear: imperfections as a form of rejection of digital makeup, functionality without disguises, frontal rejection of prefabricated templates. “We don’t need decoration, we need design that just works,” summarized the people from the U1CORE design team when analyzing one of the many tentacles of this anti-AI Slop: the brutalist minimalismwhich is the label under which this new design trend is also categorized We have philosophy. And China, no less. Some evoke the aesthetics of another architectural and decorative trend: Japanese wabi-sabiwho finds the ephemeral and the defective beautiful. Cracks in walls and objects, time-worn textures, organic asymmetry… everything that algorithmic perfection rejects, anti-AI slop highlights. Many designers have named it “post-AI visual fatigue“the feeling that has given rise to all this: a collective exhaustion in the face of designs as polished as they are sterile and devoid of personality. Who said punk? For some of us, those of us who are old dogs, this philosophy reminds us of the guidelines of the first punk, the one who created fanzines with headlines made with letters cut out of magazines. Then ethics became aesthetics, and everything was militancy of photocopying and album covers as if they were kidnapping notes; But along the way, there was also opposition to a giant. To serious media, with gray designs and content without stridency. Punk stood up to the establishment with filth and “do it yourself”. It sounds very familiar to us: AI is the new mainstream, and many are going hardcore mode. Header | Kris Shakar In Xataka | Young people have decided to stop posting (so much) on Facebook and Instagram. “AI-generated garbage” has free rein

A very rare element of the periodic table is unleashing a new geopolitical battle with China: Germanio

China has been weaving, little by little, a network of power around critical minerals: first Rare earthsafter Copper And now Germanio. Although its name barely says anything to the general public, this metal is essential for the defense industry – from the night vision systems of the fighters to the satellites – and for the optical fibers that support the Internet. Today there is almost no market, its price has been quintupled in two years and the origin of the collapse has a clear name: Beijing. The origin of the crisis. Two years ago, China announced controls At the exit of Germanio, Gallium and Antimony in response to the restrictions of the United States and the Netherlands on advanced semiconductors. However, the real blow arrived at the end of 2024: Germanio’s exports collapsed, leaving merchants without supply. Terence Bell, from Strategic Metal Investments, I recognized Financial Times That had been able to buy a gram six months. “The situation is desperate,” he said. Aaron Jerome, from Lipmann Walton & Co, described a devastated market: “Before we could buy 100 kilos; now we are lucky if we got 10, and the triple price.” And Christian Hell, from the Tradition Commercial House, added to the same medium that the demand was “for the clouds” and that he received desperate consultations of companies from the United States and Europe. The figures confirm the collapse. According to a Policy Accelator Silverado analysis cited by Financial Timesbetween January and July of this year, Germanio imports to the United States from China fell 40%. The result has been an unprecedented price escalation: just $ 1,000 in 2023 to almost $ 5,000 in September this year. This is the highest level registered since 2011. A strategic role. The importance of Germanio is not in its geological rarity, but that it is very difficult to extract, since it is obtained as a zinc and coal byproduct. In addition, its use in defense is irreplaceable for thermal image systems in fighters, drones and satellites. In the civil sector, it is used in optical fiber, solar panels and chips. “Finding substitute materials is complicated, because it would imply a complete redesign and a loss of unacceptable precision in military applications,” explained the analyst Caroline Messecar in Financial Times. For these reasons, According to estimates from the Fastmarkets agencyworld demand is around 180-200 tons per year of Germanio. One more piece of a much wider board. In Beijing they have converted critical minerals into geopolitical weapons. At the end of 2024, They prohibited export from Gallium, Antimony and Germanio to the United States, and shortly after added Scandio and Disposioessential in chips, telecommunications and storage. The strategy behind the Asian giant is to monopolize the control of the entire chain. To name a few examples, China has 4% of world copper reserves, but controls 49% of the global refining. “More than accumulating raw materials, China is building an intentional bottleneck in the supply chain,” My partner has detailed in Xataka. The same goes for the Tungsten, where it controls 83% of the world supply and tightened the export controls in February 2025, What fired prices 55%. In simple words: Beijing seeks to be essential. It controls the most valuable link – the defendant – and with it conditions global access to strategic metals of the 21st century. However, its power is not absolute: it depends on importing concentrates from countries such as Chile, Peru or Mexico. If any of those partners change position – Mexico, for example, 50% tariffs have already imposed Chinese products in 2025-, Beijing risks a cut of vital supplies. In addition, this control strategy has a price: Chinese copper foundations work with negative margins and some have had to close. A movement to counterreloj. Before the blockade, Germanio’s great consumers try to move quickly. On the one hand, in the United States, defense giant Larkheed Martin signed in August a direct agreement with the South Korea Zinc to ensure supply, something unpublished so far. Lightpath Technologies, with government support, works in optical alternatives, although its director Sam Rubin warns in ft: “No one is going to redesign an existing system until it is inevitable.” On the other hand, the options are scarce. Umicore in Belgium and Teck Resources in Canada produce some Germanio, but insufficient. Germany He already warns thatif the crisis lasts, its automotive industry could stop part of the production in a matter of weeks. The European Chamber of Commerce has even asked Beijin to release supplies for chips factories. The historical supplier, Russia, has also been out of the board. For years it was one of Germanio’s main sources for the West, thanks to its production associated with zinc and coal mining. However, international sanctions for the Ukraine War cut that flow almost completely. Moscow continues to produce, but its exports are now directed to China and countries that do not participate in the sanctions, According to FT. For the United States and Europe, that means having lost another supply route in the worst possible time, which has further reinforced Beijing’s domain. Looking to the future. In Germany, a group of researchers from the Technical University of Freiberg Work in a method surprising: extract Germanio from plants after fermentation processes for biogas. At the moment, they only achieve some milligrams per liter, but they aspire to reach a gram, which would open the door to a sustainable and local production. From anonymity to key element. Germanio has become a symbol of a new era: that of minerals as strategic weapons. As Financial Times has pointed outdemand does not stop growing while the offer narrows. And the lesson is clear: in an electrified and militarized world, who controls critical minerals will control power. Image | Freepik and Unspash Xataka | Nickel’s paradox: West needs it more than ever for electrification, but China and Indonesia have market dominance

These rabbits in a springboard are funny. And they are unleashing one of the biggest trusted crises in AI

At some point it had to happen: a wave of people wondering, and the doubt extending virally, about when the images and videos produced by artificial intelligence too realistic so that we cannot distinguish them from reality will be. And above all, what crisis of trust in what we see will take that. And the video that has unleashed the crisis is the most harmless possible: a lot of rabbits jumping in an elastic bed. Unleashed rabbits. The video is Extremely simple (Just eight seconds): A group of rabbits captured with a security camera that points, during the night, to an elastic bed, have fun taking advantage of the darkness, jumping on it. Result: more than 180 million views, and that the video is obviously false, with a two -headed rabbit at the beginning and two rodents that merge into one towards the end (although a detailed analysis by an expert will make clear the many video problems). Network movement. The video moved online, and soon jumped to other social networks. Quickly his artifice was detected (and For examplein X he was soon marked by the community’s notes), but his adorable images had more power than their failures. On the one hand, you have to have your eye trained to detect your problems, on the other the poor image quality that adds likelihood and the natural of the movements of rabbits do the rest. And above all, there is, as a huge question, his absolute lack of objective: why would someone want to deceive you with this? And that is the key. Animals in springboard. Of course, this has generated a Imitors rale (as This beareven more realistic than rabbits, although with an extra problem: the ball seems to be attached to the canvas) that have only underlined the original drama: we laughed at our elders because Ia was cheating on them, but this already affects everyone. There are those who are responding to the rabbits of the elastic bed and their power to make us doubt everything With songs (“Sometimes your love feels as real as an unknown rabbit”) or with the finding of “I am old“It is true, the video has a lot in favor to deceive us (the fixed fund makes the AI do not have to generate precisely what the IAS usually fail), but fear … is real. The crisis of trust exists. According to AI is sophisticated, we are developing, as a psychological protection, a “crisis of trust” that is already studying with figures in hand. The advice on how to combat misinformation They happen And there are those who affirm that trust is, in fact, The authentic value of the economy of artificial intelligence, above other capacities. It is no longer that AI is ending our confidence in those who have to tell the truth, Like journalistsbut it is making us doubt our perceptions. According to the daily use of AI shoots, also Our distrust does. It is paradoxical that the apparent inconsequence of a video of animals filmed with a night camera is the first notice. But somewhere you have to start. In Xataka | We have a philosophical problem with the generative AI: they are giving us the reason in everything we ask them

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.