Since the beginning of 2026, I have found it harder to make sense of the world around me.
There was no single event. The feeling comes from ordinary encounters with people, products, organizations, and the news. Many things still work, but their direction often surprises me. The way I learned to understand these things no longer explains what I see. I used to keep a collection of bookmarks for blogs I enjoyed. Some writers discussed technical problems. Others wrote about philosophy, society, or whatever had caught their interest. I returned to those sites because I wanted to know what the writers thought. I did not need them to agree with me. Most of those writers have stopped publishing. These were not blogs I read once and forgot. I followed them for years. They have not continued publishing elsewhere, as far as I have been able to find. I miss their voices, not just the subjects they covered. This seems strange to me. It has become much easier to publish a thought or share an idea. The internet holds more material than I could read in a lifetime. Yet some of the people I most wanted to read no longer write for the public. I have also noticed this silence among people of my generation. I sometimes wonder whether we have grown tired of the open internet. I do not know why those writers stopped. I cannot say that an entire generation has stopped writing. But the change is real in the small part of the internet I once knew well.
Another small puzzle comes up when I try to buy a phone. I prefer a small phone. A screen of about 4–4.5 inches feels comfortable and is easier to carry around. I can hold it securely and use it with one hand. I rarely use my phone for media consumption. These are simple requirements. Yet it is remarkably difficult to find a new phone that meets them. There are countless phones to choose from. Their cameras, displays, and processors continue to improve. But most new models follow the same direction: bigger screens and larger bodies, partly to accommodate larger batteries. I can see why many buyers prefer them. That does not make my preference disappear.
I notice smaller versions of the same problem elsewhere. I would like a headphone jack on my phone. Sometimes I want a mobile plan for calls without buying a larger data package. I would rather read a physical menu at a restaurant than open a link on my phone. None of these requests seems difficult in itself. Still, the common option increasingly seems to be something else. We can produce and distribute more things than before. Why do some ordinary choices seem harder to keep?
More of everything, less of something [“The world’s still the same. There’s just… less in it.” – Capt. Jack Sparrow]
A large number of available products does not guarantee that those products offer many different experiences. Ten manufacturers can sell ten different phones. But they may all design their products for a similar group of buyers. I can choose among those phones, yet none may meet my needs. I do not think a manufacturer must serve every preference. Small phones have design limits. A larger device can hold a larger battery and give more space for a screen. A business also has to consider development costs and expected sales. My preferred product may have too few buyers to justify those costs. I understand those reasons. What surprises me is how thoroughly some categories disappear. An option that was once ordinary becomes a special request. In some cases, it becomes impossible to buy at all. Market demand can explain the result without making it satisfying for everyone. I notice another version of this problem online. The internet contains an extraordinary range of writing, music, and video. But the material I encounter depends on the services through which I discover it. Availability and visibility are two different things. A writer can publish something without reaching readers who would value it. I have watched the internet become more visual. Images and short videos now play a large role in how people present information. That is not necessarily because every user has rejected long-form writing. Creators also respond to the formats that platforms make easier to distribute. A reader can prefer essays while receiving a steady stream of clips. I do not want to claim that recommendation systems make all good writing invisible. Nor can I attribute the disappearance of my favorite bloggers to these systems.
Still, I wonder what happens when distribution rewards one form of expression more than another. People who want an audience may change how they write or what they produce. Some may decide that the effort no longer brings enough value. Others may continue because they enjoy the work. These choices can gradually change what appears on the internet. There is another concern here. More material does not always mean more useful information. I can find an answer to almost any question in seconds. But I often spend longer trying to decide whether the answer has context or represents careful thought. A confident explanation can save time. It can also hide what the author does not know. My concern is not that everything on the internet has become worse. I still find excellent writers and thoughtful work. The difficulty is that volume, quality, and visibility do not move together. I sometimes have more material to choose from and fewer reasons to trust what appears first.
These differences may explain part of my discomfort with modern products and information. Technology expands what people can make. Markets and platforms influence what becomes common. A manufacturer follows sales data, another copies what seems to work, and soon an entire category starts looking the same. Creators respond to what gets distributed, and what gets distributed shapes what audiences discover. Each decision may be reasonable on its own. Together they produce an outcome that nobody really sat down and designed. I find that fascinating (read really concerning), and a bit difficult to accept when I am on the wrong side of it.
When appearance becomes valuable
As the internet becomes more visual, I wonder if it also changes the way we value experiences. Maybe even our psyche, though I have no way to measure that. I have noticed this across many small things. More experiences seem to come with an expectation that someone will photograph, record, or share them. I see this in festivals, weddings, travel, and ordinary social events. Sometimes the plans give as much attention to the image as to the experience. I see cafes where the decor feels like the main product and the food is almost secondary. I am not saying that is bad. It is a marked change in what a place offers. Cafes have been places to read, talk, meet people, or simply spend time. Some now seem built as much for being photographed as for being in. Perhaps that reflects what we value. Or perhaps what we see every day slowly changes what we learn to value.
I do not know whether people have changed as much as I suspect. Weddings have always involved public display. Families have long used celebrations to show status, belonging, and generosity. Portraits existed before smartphones. People did not begin caring about appearances when social media arrived. What has changed is our ability to record and distribute almost every moment. A photograph no longer needs a special occasion or professional equipment. A video can reach hundreds of people before an event has finished. Other people can respond immediately. That creates a new setting for choices that once had a smaller audience. I sometimes wonder whether the expected photograph influences the experience before it takes place. People may choose a restaurant because the room looks good in pictures. They may plan a celebration around how it will appear on a screen. That does not mean the celebration is less meaningful. But the audience outside the room becomes part of the event. The attention given to selfie cameras is one small sign of how important self-presentation has become. It does not prove that people care less about other things. Good cameras serve many practical purposes. Still, the products we develop often reflect the activities that have become important to buyers.
I see a similar tension in entertainment. Films now have access to extraordinary visual technology. Some productions can create worlds that earlier filmmakers could hardly have imagined. Yet I often find myself searching for something unfamiliar. Many films seem to return to recognizable characters, structures, and settings. Even with all the VFX and processing, I sometimes feel more realism in old Mickey Mouse cartoons or Ramayana: The Legend of Prince Rama. That film by Yugo Sako, Koichi Sasaki, and Ram Mohan came out in the early 1990s. Of course, I am not talking about photorealism. I mean something about how believable the world feels to me, even when I know every frame is drawn. I am not sure whether originality has actually declined. The films I remember from earlier decades have already passed through a strong filter. I remember the ones that mattered to me and have forgotten many ordinary releases. Today's films have not had the same time to settle in my memory. Maybe age is catching up with me, and I don't register new memories in quite the same way. Michael Stevens talks about a related idea in the Vsauce video Illusions of Time. He describes a "short-short", time passes quickly while we are absorbed, but leaves little behind when we look back. I sometimes wonder how much of my own sense of repetition comes from the work itself, and how much comes from the way I experience and remember it. And while I promised myself I would not go on another rant about favorite YouTubers, I do notice how differently creators such as CGP Grey and Veritasium fit into today's publishing rhythms. I cannot claim to know why any of them change their pace or format. Still, their work makes me think about the space available for making something carefully. Then there are the decisions made before we even get to watch a film. A large production puts a great deal of money at risk. Familiar material can be easier to finance and market. Streaming services measure attention and completion, and those measures may shape what gets funded next. My ML brain sees an exploration–exploitation problem here. If something performed well yesterday, organizations have every reason to repeat it. Do that long enough, and you may reduce the diversity of what you try. But the audience is not stationary. The highest-reward choice from a few years ago might not be the highest-reward choice today. I just find the pattern hard to ignore. A film can be technically impressive and still leave me with little to think about. Another can use simple tools and stay with me for years. I am interested in why scale, polish, and immediate engagement often become the first things we measure.
Short videos raise a similar question. Their format gives an idea very little time to earn attention. Many creators therefore begin with a striking image or statement. That is a reasonable response to the format. But some ideas need time before they become interesting. An argument can be difficult at first and useful later. I cannot (and dont want to) conclude that short videos have reduced everyone's attention span. People still read difficult books, study complex subjects, and spend hours on work they value. The more precise question concerns what our daily information systems reward. They may encourage quick reactions even when readers are capable of sustained attention. I notice this pressure in writing too. Professional messages, articles, and online posts often arrive in a clean and familiar style. AI makes that style easy to produce. A person can prepare a report or a long article with very little writing effort. The result may be clear and useful. It may also contain little of the person's own understanding. This is where appearance becomes more than an aesthetic question. We use appearance to judge effort, care, and competence. A carefully prepared document once suggested that someone had spent time making it. That was never a guarantee of good work. But the relationship between appearance and process is changing. I have caught myself reacting against polished writing almost on instinct. If something looks too clean, too evenly structured, I start distrusting it. That is probably my own defensive mechanism, and maybe I have swung too far in the other direction. Messy writing does not automatically contain an honest thought. Nor does clear writing imply that someone has none. I know that, yet the instinct is there. It worries me that I may replace one weak signal of credibility with another. I am increasingly unsure what I should trust.
The difference between doing and producing
I feel this most clearly in my own work. I have spent more than a decade of my life in machine learning. Research has always given me particular satisfaction. I start with a question that I cannot answer. I read, build an intuition, and run an experiment. The experiment often shows that I understood less than I thought. Then I return to the problem. I check the assumptions and inspect the results. Sometimes the mistake is simple. Sometimes I have to change the way I understand the problem. After several attempts, an idea finally becomes clear. That moment is difficult to replace with a result on a page. The result matters, but the work leading to it matters too. I develop intuition by seeing why an approach fails. I also learn which questions are worth asking. Some of the value comes from a process that is not visible in the final paper/report. I know how this might sound. Someone could have made the same complaint when calculators became common. Why should anyone spend time on calculations that a machine can do correctly? Removing routine work can free people to address larger problems. Research has always improved through better tools. I agree with that argument. I work on AI systems, and their progress excites me. These tools can help people write code, explore research papers, and test ideas. They can remove work that adds little understanding. They may also let more people participate in research. But not every difficult step is routine work. Sometimes the difficulty is where understanding develops. A person can memorize the rules of linear algebra without seeing what the rules mean geometrically. Then a different explanation can change their understanding. That moment has value beyond completing an exercise. For the person experiencing it, a small shift in understanding can change a lot. The same distinction applies when an AI system solves a problem for me. I might receive a correct answer without knowing why it is correct. Or I might use the answer to investigate something that was previously out of reach. These are different experiences. The tool does not determine which one I have.
The way we organize work may become increasingly important. A researcher who uses AI to test an idea can learn from the process. Another researcher may delegate nearly every difficult decision and review only the final result. Both can produce useful outputs. But they may develop different skills and have different levels of confidence in their own understanding. The distinction appears outside research. A person can enjoy learning an instrument even when recorded music is easier to obtain. Someone can cook because the activity is satisfying, not because it is the fastest way to get dinner. A writer can discover an idea by trying to express it. The finished result is not the only reason for doing the work. Of course, some people dislike these activities. A person who wants a quick answer should not be required to enjoy the process. The value of doing something depends on the person and the task. I do not want to make effort into a moral requirement. I only want to understand what changes when we remove it. I suspect that we will find new forms of satisfaction. Scientists will work on questions that previously seemed impossible. Artists will try new methods. Students may learn in ways that were unavailable before. I am excited by these possibilities. Still, I cannot treat greater output as a complete measure of progress. The experience of learning, making, and understanding also matters to me. I have not found a simple way to compare these things.
The assumptions behind our institutions
When I expand this beyond individuals, I begin thinking about institutions. They often judge people through what they produce. Universities assess assignments and examinations. Employers review written work and completed projects. Research conferences evaluate papers. These arrangements depend on assumptions about what the output tells us. Some assumptions were practical rather than deliberate. Preparing a research paper once required substantial effort from the people involved. Producing a convincing manuscript still requires care if the work is genuine. But software can now imitate much of the language and structure at a low cost. The visible form does not establish that the underlying contribution is sound. This puts pressure on systems built to examine that work. A reviewer needs time to study the claim, the method, and the evidence. A polished manuscript can be generated far faster than a careful review can be completed. The problem becomes more serious when many people submit such material. For ICLR 2027, the conference introduced new submission limits, including a cap of 20 papers per author. It also limited how many papers an author could submit when none of the coauthors met its experienced-reviewer criteria. The organizers cited strain on the review process and concerns about low-quality submissions, including AI-generated work. They were trying to protect a process that must separate valuable work from weaker work.
There are many reasons why research submissions increase. More researchers work in machine learning, and more organizations can participate. AI also changes the cost of preparing a manuscript, sometimes dramatically. I suspect that matters, but I cannot tell how much of the increase it explains. The conference policy does not establish that AI caused the surge. It does show that an existing institution has had to change its rules. This is not only a problem for science. A news article can look credible without being well checked. A resume can describe skills that a polished document does not demonstrate. A business report can present a confident conclusion while hiding weak assumptions. The old signals of effort and expertise may become less useful in some settings.
I wonder whether the difficulty is genuinely new Ann Blair studied how scholars managed abundant information before the modern internet. European scholars complained about excessive information after printing expanded. They developed indexes, reference books, and methods for selecting what to read. They did not solve the problem by reading everything. Instead, they built new methods to manage knowledge. I find the comparison useful, though it does not settle the present question. An index points to text that a reader can inspect. An AI system can create a new answer from many sources. It can also make an error that sounds like a well-supported statement. The technologies are different, even if some human problems appear familiar.
Industrialization offers another comparison. Machines changed production, but the consequences extended beyond the machines. Employers reorganized work. Workers faced new demands. Governments introduced rules that changed over time. The effect on an individual depended partly on how an organization used the new technology. That history makes me question a simple argument about AI and work. Better tools do not determine how people will use them. Organizations choose what to measure, what to reward, and what tasks to delegate. Workers have different levels of influence over those choices. The results can vary even when the same technology is available. And there is a related lesson in technical standards and international institutions. Some arrangements persist because replacing them would cost too much. Others remain because participants cannot agree on what should replace them. A rule can be useful when people first adopt it and become awkward under new conditions. It may still remain the rule. I think of the postwar Bretton Woods system, which eventually had to change as its monetary commitments became harder to maintain. International cooperation did not disappear with that arrangement. I do not want to turn this into a history lesson. What interests me is that I grew up treating many institutions as established solutions. Now I see them more as arrangements shaped by earlier constraints, compromises, and sometimes the people who had the power to decide.
Perhaps the present feels strange partly because those earlier conditions are changing. People still expect familiar institutions to perform the same functions. But the work needed to maintain them may be different. The participants may also disagree about how to respond. I do not think every institution is failing. Some have adapted remarkably well. Others will need time. The more difficult question is how we identify which assumptions still hold. That task becomes harder when different participants benefit from different outcomes.
Different experiences of the same future [Jagged intelligence will have second-order jagged societal impact.]
I am particularly optimistic about India. I see younger people attempting ambitious work in science, technology, and business. Skyroot was such a pleasant surprise to me. Some are building products for problems that received little attention before. Others want to compete directly with organizations around the world. I find that ambition encouraging. I also see a change in what seems possible. A person with access to good tools can learn, build, and test ideas at a much lower cost. That matters a great deal in India. We have spent years discussing the gap between our research ambitions and our investment in research and development. Money remains a constraint for many people with good ideas. Better tools do not erase that gap, but they can lower the cost of the first experiment. Small groups can attempt work that previously required a large organization. This does not remove the need for money, skill, or patience. But it can change where people begin.
I would not want my earlier observations to hide this excitement. Better tools can give people more freedom. They can help a student understand a difficult subject. They can let an entrepreneur build a first version of a product. They can help researchers explore questions that once required more resources. At the same time, India makes the differences between people's experiences difficult to ignore. The people building AI systems, or working closely enough with them to understand their limits, are a tiny part of the population. Many others have different access to education, reliable information, language support, and economic opportunity. A change that creates a new profession for one person may affect another without much explanation. I worry about the gap between using technology and being able to judge it. A person can use a smartphone every day without knowing how its information is selected. Someone can receive an AI-generated answer without knowing the limits of the system that produced it. Of course, nobody needs to understand neural network internals to decide whether a claim is trustworthy. I work in this field and still get things wrong. The real issue is knowing when to question an answer and having some way to check it. I am really worried on this front, especially where parts of India may be moving through a different history of information. Some communities may have had limited access to books, long-form journalism, or established publishing institutions. They now have access to phones, video platforms, and AI-generated answers. Their ways of evaluating information may develop through different habits and sources.
I do not have enough evidence to describe this as a general pattern across India. I have anecdotes, not a dataset, auto drivers in tier-2 and tier-3 cities scrolling endlessly, people over 60 getting drawn into unfamiliar online worlds, and the scams that seem to follow. But these are different problems, and I cannot explain them all through low information literacy. People of every age and background can get caught in these systems. I also do not believe that reading books is the only way to develop sound judgment. People learn through conversations, community experience, local knowledge, and many other sources. Still, the question troubles me. What does a person use to judge credibility when convincing material becomes easy to generate? A confident voice and professional formatting can make an answer seem reliable. Citations can create the impression that someone has verified a claim. Yet each of these signals can be produced without sufficient care. The problem is not limited to people with little technical experience. I see highly educated professionals accept weak arguments when the presentation is convincing. I have done this myself. Familiar language can make a claim feel true before I examine the evidence. Technology may amplify a human tendency that was already there. Some people will have the time and resources to verify important claims. They can consult experts, compare sources, or test an answer directly. Others may have less time and fewer options. The difference between receiving information and being able to evaluate it could become significant. I do not know how quickly our education and public institutions can respond.
This is why my optimism about India exists beside uncertainty. I can admire the ambition of people building new systems. I can also worry about people who will live with the consequences but cannot influence the decisions. Both observations seem reasonable to me. Neither removes the other. I do not want to reduce India to a story of technological inequality. The country has many forms of knowledge, initiative, and adaptation that do not appear in discussions about AI. Nor do I think the future will move in one direction for everyone. Perhaps the world has changed faster than the way I understand it. Or perhaps I never understood some of its rules as well as I thought. I still don't know.
Come gather 'round, people, wherever you roam
And admit that the waters around you have grown
And accept it that soon you'll be drenched to the bone
If your time to you is worth saving
And you better start swimmin' or you'll sink like a stone
For the times, they are a-changin'
— Bob Dylan, The Times They Are A-Changin’