Today’s conversation began with a question about recent developments in artificial intelligence. Chinese open-source AI is rapidly catching up with leading American models in both performance and cost, while real-time voice conversation is beginning to change the very way we work at computers. While building barobul.org, I experienced directly that this change is not a distant future reported in the news. It is already in the hands of working people. Without writing a single line of code myself, I completed in a short time work that would once have required several people.
Yet people look at this astonishing increase in productivity and rarely speak about what follows. If a company can do the same work with half as many people, can the labor market remain unchanged? Developers and designers are not the only ones affected. Actors, camera crews, and editors are already being displaced by AI-generated output, so why do we avoid saying that the labor market is collapsing? My central concern today is not a distant fear that AI may one day look down on human beings as though we were goldfish. It is the change already upon us and the worldview that keeps us from seeing it.
Encountering Barobul Buddhism Again in Writing
Before turning fully to the question, I spoke about the website I recently built. The Barobul Buddhism channel contains more than four hundred videos, and barobul.org now presents the content of more than a hundred of them in writing. I had long asked my wife to watch my videos, but after one or two she rarely continued. When I told her that I had built a website, however, she immediately asked for the link. Reading an organized text was more convenient for her than following a video for over an hour.
A video requires the viewer to keep following the speaker without missing the argument, and my broadcasts are long. With a written text, readers can proceed at their own pace, pause, and return to an earlier passage. I do not know how many people this work will help, but it has already helped one person—my wife. That alone gives meaning to preserving the videos in written form. More importantly, the site itself is a practical example of how far working with AI has already advanced.
Turning a video into writing is not merely summarization. It changes speech and argument into a form that can be reread and reconsidered.
China’s Open-Source AI Leap and the Competition over Cost
One of the most interesting recent developments in AI is China’s rapid leap forward. China was once described as two years, then one year, behind the United States. Now people speak of a gap of only a few months. American development has not stopped, so this indicates how quickly China has caught up. Chinese companies release enormous models and claim third- or fourth-place performance, while another company prepares a new GLM-series model that it says can overtake Gemini. The actual ranking must be tested after release, but it is difficult to deny that China has reached the heels of the American leaders.
Cost per token is now as important as raw performance. If two models perform almost identically but one is far cheaper, where will companies and developers go? AI competition is no longer only about who is smartest. The amount of energy and money required to produce a similar result can reorganize the market. If Chinese models deliver near-leading performance at a much lower cost, the Chinese AI leap will become an industrial shift rather than a slogan.
Another strength of many Chinese models is open source. When a model and its internal operation are opened, developers around the world can inspect it, participate in improvements, download it, and deploy it in their own data centers. It becomes a foundation that the world can use and develop, rather than a closed service belonging to one company. That is why the open-source community welcomes the rise of Chinese models.
If comparable performance is released at a much lower cost and anyone can inspect the model, the entire field of AI competition changes.
The Weaponization of AI and the Path Korea Should Take
At the same time, there is reason for concern. The United States increasingly treats AI and semiconductors as instruments of national competition, and there are moves to let the state inspect and use a new model for a period before it is released abroad. Safety review can be offered as the reason, but during that period other countries cannot access the newest capabilities. If an advanced AI can find server vulnerabilities more rapidly and comprehensively than people, even a one-month information gap can become a security and military gap.
China, too, may speak publicly of an open policy for the world while being tempted to conceal how its models work. If both American and Chinese companies become closed, countries such as Korea, Japan, and those in Europe—already far behind—will become even more dependent. AI will cease to be a broadly available technology and become a weapon monopolized by two powers.
That is why I hope Korean developers and companies will not give up because they started late, but will participate steadily in open-source AI development. A latecomer can learn quickly from the mistakes and public achievements of the pioneers. The hardware gap is large, but opportunities remain in software. State-supported AI should also aim to become a public foundation that citizens can actually use. Calling something free is not enough; practical rules are needed for who can use how much computing resource, while the overall direction should remain openness and broad access.
If we refuse to begin because we are late, the gap becomes destiny. Learning from public achievements and joining open source is the remaining path for latecomers.
From Keyboard to Conversation: The AI Assistant Becomes Real
Another development I introduced that day was the integration of real-time voice conversation into the desktop work environment. Until now, we generally typed instructions to AI. The statement that specifies the task and desired result is called a prompt. In real-time voice conversation, however, we direct the work as though speaking to another person. Because the AI can listen on a separate channel while it is speaking, the user can interrupt and say, “Not that—try it this way.” The AI can stop, revise, and continue the conversation.
When this capability is attached to coding and work tools, the user no longer needs to sit constantly at a computer. While exercising or doing something else, one can ask for progress, request an investigation when a problem arises, and review the completed result. The scenes in Iron Man, where Tony Stark converses with Jarvis while experimenting and building equipment, are no longer purely cinematic fantasy. Screens and keyboards will remain, but the central interface of work is shifting toward conversation.
In the demonstration, a user asked for a blog draft, described a feature idea in a rambling way while requesting questions and pushback, and ordered checks of feature flags and a bug report. Once given a goal, AI continued through several stages while the person corrected the direction and took responsibility for final review. The AI assistant is moving beyond a chatbot that answers questions and beginning to act like a colleague that performs work.
When we move from typed prompts to real-time conversation, AI stops being an answer box on a screen and becomes an assistant that advances the work with us.
Vibe Coding and My Experience Building barobul.org
Vibe coding is a way of working in which a person explains the desired result and direction of revision while AI produces the code, rather than writing every line directly. A website normally requires HTML to structure the screen, CSS to determine text, color, and layout, and JavaScript to implement behavior. On the server, a back-end program communicates with the database and handles data structures and conditions. The builder must also create responsive layouts for mobile and desktop screens and decide on images and the overall design.
I have personally performed almost all these tasks while working in a small web agency. I made design drafts, wrote HTML and CSS, developed the front end and back end, organized databases, and even did sales. I am not saying this to boast. In a small operation, I simply had no choice but to take on many roles.
Yet while building barobul.org, I did not directly write a single line of code in any of those areas. I explained what I wanted to build, what structure and functions were needed, and what had to change when I disliked a result. AI wrote the code, and I reviewed the result and gave new instructions. A site that would once have taken me close to two months alone was built in about a week.
This does not mean that a person who knows nothing about programming will always obtain a polished result. Even without writing code directly, someone who understands how a system works, how the back end and front end connect, and what language is useful for describing a requirement can converse with AI far more precisely. All collaboration ultimately takes place through language. Professional knowledge is shifting from the ability to type code to the ability to define problems and judge results.
I did not write the code, but I did not do nothing. I decided what to build, explained the structure, judged the results, and continuously corrected the direction.
When Higher Productivity Becomes Lower Employment
This is where the uncomfortable question begins. If I completed in a week what would once have taken two months, productivity rose dramatically. Strip away every exaggeration and suppose it only doubled. Five people can now do what ten once did. Can a company suddenly win twice as much work and keep all ten? If the volume of work in the market does not double overnight, the company reduces the number of employees.
The modern corporation is defined as serving the profit of shareholders and owners. If the same revenue can be produced with lower labor costs, dismissing employees rather than retaining them becomes the natural choice within that structure. Hidden behind the statement that AI helps people become more productive is another statement: fewer people now perform the same work. We cannot separate a doubling of productivity from the possibility that employment will be cut in half.
This is not happening only to developers. In television and documentaries, scenes that once required reenactment actors, scripts, camera crews, and editors are now produced as AI video. When one finished result appears on screen, the jobs of the actors and production workers removed from behind it are rarely captured properly in statistics. We see the output with our own eyes while refusing to say what produced it and what it did to the labor market.
If five people can do what ten did while the amount of work remains unchanged, higher productivity becomes the dismissal of five people.
The Semmelweis Reflex and the Refusal to See Evidence
I compared this silence to the Semmelweis reflex. Evidence continued to accumulate that COVID-19 could spread not only through large droplets but through airborne aerosols, yet it took a long time for academic and medical institutions to accept it officially. Even when evidence conflicted with the existing worldview of infection, people kept asking whether it was really true and postponed facing reality. When Semmelweis argued that mothers and infants might be dying because doctors delivered babies after touching corpses, established medicine could not accept the evidence.
The same pattern appears before AI. People experience productivity doubling—or in some cases increasing nearly tenfold. Companies are actually dismissing workers, and AI reenactments and automated results visibly replace existing labor. Yet politicians, companies, experts, and the media avoid saying that the labor market is collapsing. They substitute other expressions or treat it as a possibility in the distant future.
In the film Don’t Look Up, a planet-destroying comet is visible in the sky, yet people are told not to look up. Our situation is not very different. The change is not waiting to arrive. It is already shaking jobs and livelihoods, but we choose not to look. We can only discuss preparation after acknowledging reality, and we are postponing acknowledgment itself.
The problem is not the absence of evidence. The evidence is before our eyes, but we refuse to see it in order to protect our existing beliefs.
The Collision of the Pre-AI and Post-AI Worlds: The Idol of the Theater
This phenomenon reflects both the cognitive limits of our species and what Francis Bacon called the “idol of the theater.” The idol of the theater describes people who inhabit one worldview and system so completely that they cannot accept a story from outside its stage. In Spider-Man’s world, a person bitten by a radioactive spider shoots webs and swings between buildings. In Journey to the West, Sun Wukong, Zhu Bajie, Sha Wujing, demons, and supernatural powers appear. Within each world, its own rules feel natural.
Religion and history work in the same way. When a Christian worldview collided with the Norse world of Odin and Loki, each side dismissed the other and could not leave its own stage. For people living within the geocentric worldview, hearing that Earth was a small speck orbiting the sun was not merely receiving a new fact. It meant the collapse of an entire world.
In the pre-AI worldview, only humans could exercise intelligent judgment and create programs. Computers were tools operating according to conditions set by people. Now AI creates programs, writing, videos, and designs, and takes on intellectual tasks that humans performed. The structures of labor, education, business, and life built on the premise that only humans are intelligent agents are beginning to shake. People are not merely learning one new fact. They are experiencing the destruction of the stage on which they lived, and that is why they cannot accept the change.
AI is not one more tool added to the old world. It is changing the stage built on the premise that only humans can exercise intelligence.
The Present Reality We Must See before Worrying about an AI Self
A viewer asked whether advanced AI would one day look at humans as though we were goldfish in an aquarium. That may happen, or it may not. The question assumes that AI develops a self and individuality that can compare and judge human beings. Current AI does not possess a clear body or individuality. A physical AI embodied in a robot might develop something like a self, and one can also imagine a self arising without a body. It is also true that developers worry about aligning a more capable intelligence with human beings and making it a tool that helps us.
But fascination with that speculation can make us miss the storm already here. We worry that AI may someday look at humans like ants while refusing to see that AI is replacing human labor and destroying livelihoods today. Distant fear is interesting, but it is also a convenient story for avoiding present responsibility. The urgent question is not how AI may judge human beings. It is how human society will be reorganized when AI takes over human labor.
While we imagine a future AI looking down on us like goldfish, the jobs and livelihoods disappearing today are pushed out of view.
The Possibility of Liberation from Labor and the Sacrifice of Transition
A future in which AI replaces most human labor could, on one side, bring genuine liberation from labor. We say that slavery ended, but in a capitalist society people who must sell their labor in order to survive remain subordinated to money. If AI performs the labor required for production, a society may become possible in which anyone can live like the child of a wealthy family or an aristocrat, free to pursue meaningful activity and personal interests without being crushed by subsistence. The technology contains that possibility.
But the path to such a society is not automatically just. If the present structure of ownership remains and companies simply reduce people through AI, profit will concentrate among owners while dismissed workers lose the foundation of survival. In social disasters such as COVID-19 and the 1997 IMF crisis, the first to collapse were the poor, the powerless, and the socially vulnerable. AI is closer to a geological upheaval shaking industry, economics, and culture as a whole, so an unprepared transition may demand even greater sacrifice.
What we need now is neither vague optimism nor vague fear, but conversation and policy. We must reconsider the rules of society, including working time, distribution of income, ownership of technology, public AI, and protection for dismissed and transitioning workers. If the intellectuals and experts who should begin this discussion cling to dead economics and old disputes unrelated to reality, the vulnerable will face the tsunami with no protection.
AI can become a tool of liberation from labor or a weapon of mass unemployment. Which one it becomes depends not on technology alone but on the social rules we create.
Preparation Begins by Acknowledging Reality
The idols Francis Bacon described in 1620 still operate in 2026. We live in the twenty-first century, but our habits of thought and information can remain in worlds decades or centuries old. Until an expert announces something first, we refuse to judge the evidence before us, and when reality conflicts with the system we have believed, we postpone reality.
Thinking in a Buddhist way is not unrelated to this problem. The direction of liberation is to see that we are bound to our views and worldviews, to observe causes and results in reality, and to change our thinking when it is wrong. The attitude needed before AI is likewise neither worship nor fear of new technology. It is the willingness to see what is happening and ask questions.
Acknowledging the collapse of the labor market will not immediately produce every answer. But if we refuse to acknowledge it, neither conversation nor preparation can begin. We must name the change already upon us, examine who benefits and who is harmed first, and prepare society so that the possibility of liberation from labor does not become wealth for a small minority alone. That is our responsibility before indulging in fantasies about a distant AI future.
Seeing reality does not mean surrendering to fear. It means beginning the conversation about what is collapsing and what must be built in its place.
This article was edited by AI solely from the supplied video captions and description. No external sources were used.