feel

There are people who feel that the best AI becomes silly and vague over time. Is more than a sensation

by

A new one comes out artificial intelligence (AI) To the market, social networks and specialized communities hallucinate with their new capabilities, and at the same time, a cycle where users who know the models begin to feel disappointment begins. They begin to see how what until yesterday achieved without problems via Chatbot or API, today stays in a vague attempt. There is a part of sensations, and one real. “Super broken” models. When he launched, Gemini 2.5 PRO reaped huge praise in social networks. The model was Very fastof the cheapest, had A context window huge and it was A beast in programming. However, for a few weeks, comments have emerged in communities such as Reddit that describe a “unusable” model. A model, which as described worked incredibly well between March and June, but that after using now at the end of July it released “absolute nonsense“. Showed a conversation with Gemini summarized by the assistant in which he did not stop recognizing errors. Other users also show Examples of behaviors annoying as Do not finish answers. They are only recent examples of Google’s AI, but even models as praised as Claude have received at different times criticism similareven recently with Claude Code. Suspicion. Many of the users who have criticized the different models speak of cut models: “My assumption is that they reduced the size of the model,” said A Claude 3.5 user in Hacker News. The suspicion is that, over time, and at times of maximum demand, companies begin to use distilled versions of AI models that are not so intelligent, because they have less dedicated resources to respond to the indications. Ian Nuttal developer too He observed Claude Code degradationand claimed that he would pay to have a good version that would never be reduced or degraded at peak hours. Alex Finn, also developer, expressed Equally frustration: “This happened to me with all the IA programming tools that I have used.” It’s not just a sensation. In 2023, many users felt that GPT-4, Openai’s most advanced model at that time, was becoming silly. The company claimed that contrary to what the community denounced, they made each new version “smarter than the previous one.” However, a Paper Academic It ended with speculation: experts from Berkeley and Stanford checked a spectacular precision drop of GPT-4 among its variants in March and June 2023. In programming, for example “the percentage of generated responses that are directly executable was reduced from 52.0 % in March to 10.0 % in June”. Others statistical studies At the end of 2023 they also showed a significant loss of quality between the December and May model. Openai and Anthropic confirmed problems. In December 2023, OpenAI recognized that they had received the feedback on the assistant becoming more vague. They claimed that they had not updated the model from a month earlier, and that it was not intentional, recognizing the problem and explaining that “the behavior of the model could be unpredictable.” Some users came to devise (and achieve, according to their experience) Methods to encourage the model to do betterlike the surprising promise to give a tip or explain to the chatbot that they had no fingers to write the code. More recently, Anthropic acknowledged To TechCrunch have problems in Claude Code, as slower response times, before complaints of users of having limited use without having affirmed. Users who previously performed tasks normally and now could not progress. In Xataka | I have tried day, the browser that replaces ARC and bets everything to AI. It hasn’t come out as expected

If you also feel that you do not have so much time to read as you would like, there is a solution: the “fast reading”

by

I confess that Bill Gates gives me some healthy envy. Not just to have A current account which looks like a phone number with so many figures, but for its ability to Read more than 50 books a year. I have tried and it has been impossible for me. But Do not shoot the towel. Scientific evidence revealsthat there is a balance point between the speed of reading and the ability to understand and retention of the content that is being read. However, with adequate practice, that balance can be improved by accelerating the reading speed without compromising Reading understanding. Research Like those of the University of Guayaquil, they conclude that the reading speed can increase significantly with the practice and use of appropriate techniques. An average reader reads between 200 and 400 words per minute, while through rapid reading training, that speed can reach speeds of up to 1,000 or 1,700 words per minute. Strike a balance Fast reading is especially useful for processing long texts superficially, being useful for obtaining general ideas or punctual information. It is not recommended in contexts where a deep understanding, detailed analysis or memorization of the content is required, but when it comes to seeking specific information or in texts that address subjects of which previous knowledge is already had. The main ones involved in a slow reading speed are: Subvocalization: It is the habit of mentally pronouncing the words that are read. That limits the reading speed since the visual image of the word that interprets the brain is vocalized as if you are reading aloud, undergoing the need to vocalize each of the words and not limit itself to the compression of it. Word reading by word: The global understanding of the text decreases. Regression: reread several times or have to search in the next line of text as it breaks the reading flow. Low concentration: It affects both speed and retention. Quick Reading Techniques Science reveals that this speed increase is not a magical process, but the result of the conscious training of eye movements, vocabulary expansion and the use of strategies to improve global understanding, and not only the amount of words read per minute. 1. Fragmentation This technique is to avoid reading word by word and assimilating the Words by groups. For example, to start training, you can group the words of the phrases of two by two and increase the number progressively With practice. At first it may be a bit strange, but the brain recognizes the words for its morphology, which explains, for example, that you Cberreo Pdeue Read Etse Txeto. If you read it Word A Word would cost you more than in block because a word contextualizes its adjacent. This principle makes your view jump between words, not word in word during reading. The usual thing is to start reading like this:In-a-place-of-the-scan-of-whose-name-no-quiero-alleging But with a little practice, we soon start reading like this: ANDn a place-of the stain-of whose name-I do not want to remember With proper training, the number of “pauses” is reduced: In a place in the stain-of whose name I do not want to remember. 2. Use a visual guide The rupture of the reading flow is one of the most frequent causes for reading the reading. This usually happens when you have difficulty finding the following text line. The solution: as simple as using the finger to indicate the start of the next text line or an equivalent visual guide. On screens you can use the mouse pointer as a guide. If you prefer Read on mobiles or tablets, the edge the screen or any other reference can contribute to the view not to hesitate when looking for the next line and the reading inertia is maintained. 3. Be faster than voice If when you read you listen to the words in your brain, you are not reading quick enough. As we have said before, this phenomenon is known as subvocalization. To avoid this, it will be necessary to train the reading speed trying to read faster than that inner voice is able to vocalize. In doing so, it will be forcing the brain to prioritize visual processing instead of the “auditory” of your internal voice, thus improving the reading speed. 4. Infinite look Like any other muscle, ocular muscles should also be exercised to improve reading speed. One way to improve your performance is to practice infinity to get a visual sweep of the most efficient text. This technique consists in drawing the infinity symbol (∞) on the text block, so that the look runs in a more global way, not line by line, and improves the capture of words by blocks. In Xataka | The productivity books that have helped me the most: the recommendations of the Xataka editors Image | Unspash (Thought catog, Eliott Reyna)

“I feel responsible for my team and my clients”

by

According to The records of LayoffsMicrosoft has fired more than 6,000 employees in what we had of 2025. One of those employees was Freddy Kristiansen59, fired from the Microsoft subsidiary in Denmark in May 2025 after a 23 -year career in the company as a product manager. However, although officially it is no longer part of the company, Kristiansen Keep going to the office. “It may sound rare to appear in the office after being fired, but I still feel committed to the products, with the people who use them and with my colleagues,” he explains In an interview for Business Insider. A link that goes beyond the contract Its history began in 2002, when it was hired by Navision, a company that shortly after It was acquired by Microsoft. What thought it would be a temporary job finally became a professional career of more than two decades. “I thought: ‘Maybe this can be something in the long term.’ And so,” Kristiansen recalled. Although the Danish law he grants him A period of six months and three months of compensation after the dismissal for being an employee more than nine years old, Kristiansen affirms that his motivation to continue going to the office It is not economical. Microsoft relieved him of all his functions, but due to Danish legislation, Kristiansen is still an employee of Microsoft until November. Instead of using that time to look for a new job (What is designed for what is designed), the product manager continues to attend his office daily. His last great project was Somethinga tool for github. “I feel a responsibility not only to maintain the product, but also to reassure customers to know that they are in good hands,” he told the American environment. “If you need my help, I will answer questions, I will share advice or whatever necessary. There is no reason to stop doing it,” he says. Actually, they have done a favor The employee’s reaction can be surprising for continuing to go to the company Instead of racing in angerbut the truth is that Kristianse’s dismissal came at a time when new personal and professional challenges had already begun. “I thought I could feel annoying, but it really was a relief,” he confesses. While some of his companions were worried about Kristianse’s future, he saw an opportunity to lower the rhythm after two decades at the foot of the cannon. In fact, during The layoffs of 2023the veteran employee presented an anonymous consultation about the possibility of asking for a voluntary dismissal. “I never did,” he clarified. “During the last decade I have worked very hard and I have thrown many hours. But now I am no longer interested in working 60 hours a week. It seemed to me that it was the right time to finally chase my dream long postponed to work in my own terms,” ​​he said. As published in Your personal blogKristiansen focuses on designing his own business plan with the intention of offering consulting services such as CTO, management of projects and motivational talks, while Enjoy more free time And he takes up healthy habits. “From the dismissals, I repeat that there is no evil that for good does not come.” Respect for the fruit of his work Despite the dismissal, Kristiansen continues Share your experience and the decisions that marked his career. “No one forced me to work 20 hours on weekends or strive so much, but I did it because I felt it was the right thing. I did it because I really felt a connection with our partners, our clients and my colleagues. And, honestly, I still feel it,” says the former employee. Its involvement With customers, partners and products it remains so strong that, although its contract ends, respect for everything that it built during its years in Microsoft remains intact. At least until November, when you will have to deliver your corporate laptop and credential. In Xataka | A company hired a worker, dismissed the boss who selected him and forgot about him: he has been charging without working seven months Image | LinkedInUnspash (Sam Torres, Valent lau)

Figure 02 has worked only for an hour. The disturbing thing is that your brain already remembers and your hands “feel”

by

You can disagree. You may think that a lot is still missing. But it is difficult to ignore the direction in which we advance: one in which Humanoid robots They will be part of our daily lives. When technology is mature, when costs cease to be an obstacle, the decision will no longer be technical. It will be ethical. How far can the robots arrive? And who will put the limits? Imagine the scenarios. Robots turned into soldiers, deployed in conflicts, alone or next to human troops. Robots as emotional companions, not only for older people, but for anyone who lives alone. Robots that clean, cook, organize. Robots in factories, tireless, constant, replacing tasks that we did before. The surprising thing is that all this is already beginning to happen. Today, the robot Figure 02, that since last year he works in a BMW factorycan operate autonomously, classify objects and do it with precision thanks to what Its developers call “Touch” and “short -term memory”. In 2018, ‘Detroit: Become Human‘We thought about science fiction. Even then, not its creator, David CageI would have opted that in 2025 we would be seeing something like that in the real world. But it has happened. And the most fascinating thing is that Figure does not walk alone. Behind his gestures, there is something else. A neuronal network that, as we will see soon, is the true protagonist. Helix, the ‘brain’ behind Figure 02 That neuronal network has its own name: Helix. It is the brain behind the robot. The person responsible for their movements look more and more natural. And what has achieved In just three months In a logistics environment it is difficult to ignore. Helix has not only learned to manipulate objects with skill. Has learned to Understand the context. To adapt. To act as if remembered what he did before. And the most surprising: to make it better and better, faster and less mistakes. At first, Helix faced a simpler catalog. But little by little he has learned to work with a greater variety of packages: From rigid boxes to Soft envelopes or deformable plastic bags, which are much more difficult to hold and position. Some wrinkle, others bend, others slide easily. For a robot, that is a real challenge. And yet, Helix adapts. Adjust the way in which each object grabs, changes its strategy if the package is flatter or softer, even knows when it is convenient to take a small turn or use a more precise type of clamp. He does everything on the flight, without anyone telling him what kind of package he has in front. The robot knows when it is convenient to take a small turn or use a more precise type of clamp That behavior has not been programmed line per line. Has learned it by observing. In total, he was trained with 60 hours of human demonstrations. And with each new example, it was better understood what to do in each case. Thanks to that, the average time it takes to process a package has been down significantly. But Helix is ​​not just fast. It is also necessary. The labels, which were previously oriented before, now 94.4 % of the time are correctly positioned. As? Because the robot has learned, among other things, Small wrinkled envelopes Before trying to scan. A slight pressure on the plastic is enough for the barcode to be visible. It is a minimum gesture, but very revealing. That learning relies on something fundamental: memory. Helix has a vision system that not only analyzes what he sees at the time, but remembers what he saw a few seconds ago. It is as if I had a Short -term visual memory. Thanks to her, you can make smarter decisions. For example, if it detects that a package was already rotated at a certain angle, it does not turn it again. If you did not find the label at first, remember where it could be and prove in that direction. Another key change has been to incorporate the history of its own movements. Before, each action was an isolated fragment: see, act, see again. Now, Helix Remember in what position your arms werehis torso or his head makes a moment, which allows him to move more fluidly. If something leaves the plan, if the package slides or the grip is not perfect, it can correct in real time without starting from scratch. And the most recent: The sense of touch. Not a human touch, of course, but a way to detect how much pressure is exerting on an object. That strength feedback allows Helix Pause to feel resistance, or adjust it if the object weighs less than expected. Thanks to this, you can manipulate more carefully and adapt to differences in weight or rigidity. And this is not theory. Figure 02 has already been put to the test in real conditions. During an hour session, The robot operated without interruptions In a logistics environment, classifying packages autonomously while using all its new capacities: its sense of touch, its short -term visual memory, its ability to correct errors on the march. A continuous demonstration of fluid work, precise, without direct intervention. The images of that session, which we have included in the video that accompanies this article, show more than a robot running. They show a neuronal network facing the chaos of the physical world and leaving airy. They show Figure 02 making decisions, adapting, acting with a level of autonomy that until very recently would have seemed science fiction. Tesla has a OptimusBoston Dynamics to AtlasAgility Robotics A Digit. All compete to lead the career of the Humanoid robots. But in parallel, Figure 02 advances through a lane without events, without background music, but with results. With the passage of time we will see how all this evolves. Images | Figure ai In Xataka | Google does not want AI to stay on the screen: Gemini Robotics is its plan … Read more

Students feel that Pau is the opposite

by

This year, thousands of high school students face A historical selectivity. For the first time, the University Access Test (PAU) will have the same structure throughout the country, marking A before and after in the test Access. These changes are generating a lot of expectation and anxiety among students, extending the opinion that changes They have increased difficulty With regard to previous calls, in a demanding evidence. Uncertainty about How to develop The new structure of Pau is one of the main concerns among students, and the fear of the unknown has become the main protagonist of the classrooms a few days after the call. Unification and fears of the new exam model Pau 2025 is a change of structure in the access test that seeks Unify the exam model In all autonomous communities, with the objective of make more fair Access to the university, regardless of the community where the test is carried out. According to Diana Morant, Minister of Science, Innovation and Universities “is more agile, modern and adapts to society with the maximum academic rigor and guaranteeing the equal opportunities of all students in Spain.” Teachers have expressed concern that, although the exam format has been homogenized, the question of the question of the difficulty according to the community Autonomous, which could continue to generate inequalities. “The competences are transferred to the autonomous communities, and each community fixes a part of its own agenda. This makes a unique model very difficult to implement. (…) The test is organized decentralized, while the admission is unique. Even if you have to ask for a place in each university district, your note serves to enter any university of Spain,” He pointed out Javier Oubiña, Student Vice Chancellor and Employability of the Autonomous University of Madrid. However, one of the main fears of students is that this new format Increase the difficulty of the test and cause a generalized decrease in the notes. That could make access to races with the highest cut -off note and leave them in disadvantages with last year’s students. “If the notes go down a lot, it can put us at a disadvantage to enter the most demanded careers, since it is likely that those who presented the previous years have higher notes than what we are going to have. That, and the increase in the difficulty of the questions, is what overwhelms us the people of my course I have around me,” assured to The country Julia, a high school that prepares for the PAU. As He pointed out Carla Mora, 2nd high school student in Almería to Almería Diario “The changes should have already been made for students who are in 1st Baccalaureate this year and not for us that so far we have been working in a different way the test, that is, with the previous model. The novelties have been introduced to us suddenly.” Main changes in Pau 2025 One of the main News of the PAU 2025 It is the disappearance of the possibility of choosing between several optional questions. Now, all the questions included in the test are mandatory, although it will be possible to choose some questions or sections within the exercise itself. From the Ministry of Education it is warned that this “optionality will not imply being able to study less a agenda”, which forces students to prepare the complete content of each subject, and not to do without it as allowed the previous model. Another of the novelties that are worrying students is the change in the format of the questions. The new exam will combine exercises with closed tests type test, semi -constructed questions (such as filling spaces) and, the most controversial, open answers in which the student must develop the resolution process. In the latter it has been sought a more competence approach. That is, the student has understood the concept and not only memorized. They seek to evaluate creativity, critical thinking, reflection and maturity in resolution. With the arrival of the open format of the questions, the correction criteria are also modified. In them, not only will the correction of the response be assessed, but also coherence, cohesion, grammatical, lexical and spelling correction, as well as the presentation. Spelling failures will mean a 10% penalty of the score, according to the new guidelines. Both the fact of forcing students to prepare the entire age The most difficult in recent years. What has not changed in the 2025 PAU Despite the changes, some fundamental aspects of the PAU in 2025 remain the same as in previous calls. The test continues to be divided into two parts: a basic and mandatory, in which it is necessary to obtain at least 5 to access the university; and another volunteer, which allows you to add up to 4 additional points to try to improve the final grade. According What is indicated by the Ministrythe subjects of the mandatory part They are Castilian and Literature II language, foreign language II, History of Spain or Philosophy (to choose), and Co -official Language and Literature II in communities with their own language. The voluntary part includes at least two subjects of the Baccalaureate, History or Philosophy (the one that I did not choose for the mandatory phase), and a second foreign language (English, French, German, Italian or Portuguese). The final grade calculation system has not changed: 60% of the final grade corresponds to the average high school grade and the remaining 40% is provided by the note obtained in the PAU. That is, if in Baccalaureate an average grade of 7.5 has been obtained and in the Pau an 8, the final access note will be a 7.7 (Baccalaureate note x 0.6 + Pau note x 0.4). The time to perform each exam is maintained in 90 minutes, with 30 minutes of rest between tests. Students with educational support needs will have more time. In Xataka | For thousands of Spanish students, the challenge is not to overcome Pau and … Read more

One UI 7 on the Samsung Galaxy S25 has achieved its goal with AI: to finally feel the need to use it

by

I don’t hide, when I tried Samsung AI Last year I wasn’t too surprised. I found them to work quite interesting, useful in specific circumstances, but very far from the “AI Phone” they wanted to propose. This year, things have changed. And they have done it in the most natural way possible: with slight refinements. I’m going to tell you why Samsung’s AI in the Samsung Galaxy S25, S25+ and S25 Ultra It is in the sweet spot that it deserves, what makes this generation so special and why now I am convinced that it is a characteristic to use perhaps not on a daily basis, but often. We are waiting for you in Territorio S Do you like Samsung? We have a new media specialized in Samsung and its products with tricks, offers, tutorials and all the news about the brand. Go to S Territory Call recording with transcription. One of the functions most requested by xatakeros has to do precisely with call recording. There are hardly any phones that can record them by default, and you have to resort to third-party apps that don’t always work well. In the case of the Samsung Galaxy S25, it is not only capable of recording calls but, once made, we will obtain a voice-to-text transcription using artificial intelligence. It works practically perfectly, and it is a function that has pleasantly surprised me. In Xataka 2025 is the year of AI in televisions: Samsung brings Galaxy functions to the new Neo QLED and OLED Gemini as a native assistant. Bixby is still alive in One UI 7, but Gemini takes the spotlight. This is the inevitable replacement for the Google Assistant, and is capable of generating images using prompts, searching for information on the Internet and, ultimately, offering an alternative to GPT. Samsung continues to partner with Google so that the AI ​​functions that we would see in a Pixel also shine on its devices, and this is something to be appreciated. A brutal image selector. The new Samsungs go one step beyond “Circle to Search”. We can select any part of the screen, whatever we want, and the AI ​​will analyze what can be done with it. Convert a video clip to GIF, perform generative editing on a photograph, analyze texts… Any element on our screen can give us play if we give AI a chance, something that can be especially useful when they send us documents, photographs or any multimedia content. Audio removal. Without a doubt, the feature that has surprised me the most in these Samsungs is the elimination of background noise, wind and music. In the demonstration that I was able to access, the scenario was that of a person playing the piano in the middle of the street. The AI ​​was perfectly capable of isolating background noise, wind, and amplifying music. Taking into account that most of my daily recordings are made with my mobile phone, having this feature is almost magical. In short, Samsung has refined what was already working well (image generation, translations, text selections, etc.), and has implemented small touches that are very useful for everyday use. Beyond allowing you to delete elements from images and make translations (on which most of its rivals’ AI functions are based), Samsung’s interpretation (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 One UI 7 on the Samsung Galaxy S25 has achieved its goal with AI: to finally feel the need to use it was originally published in Xataka by Ricardo Aguilar .

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.