When AI rewrites the rules: five uncomfortable truths about our future

I think many people are still looking at AI too softly. We talk about productivity, chatbots, virtual assistants, faster writing, faster coding, faster research. All of that is true, but it is not the full story.

The more important shift sits underneath: AI is separating economic value from human labor. Before, if you wanted intellectual output, you needed a person who had studied for years, built experience, and then sat down to do the work. Now a large portion of that output can move through models, APIs, data centers, and workflows.

When value no longer flows through labor in the old way, society has to rewrite how rewards are distributed. This is not a normal tool cycle. It is a new paradigm.

If we keep treating AI like a better Excel file or a smarter intern, we are already standing in the wrong place.

The first uncomfortable truth is the expert paradox. The twentieth century trained us to believe that harder credentials meant safer lives. Medicine, law, engineering, finance, consulting, software: these were the routes into the upper layer of society.

AI is hitting exactly the standardized parts of those professions. Assisted diagnosis, legal document review, basic coding, data analysis, report writing, desk research. Work that once required years of training is being converted into workflow.

The line “studying medicine is worse than selling ketchup bread” sounds like a bitter joke, but the question behind it is real: what happens to the ROI of traditional education when a large share of professional output becomes commoditized? Meanwhile, very ordinary work that requires physical presence, real emotion, direct trust, or craft does not disappear as quickly as people expect. Someone selling good food, understanding customers, choosing a good location, and running operations well may have more stable cashflow than a young graduate carrying an expensive credential while their entry-level skills are being copied by AI.

This is not an argument against studying. It is a reminder that a degree is no longer an automatic moat.

The second truth: AI may not fire old workers first. It may block new workers first. Reports such as the Anthropic Economic Index, along with what we can already observe in hiring markets, point to an uncomfortable pattern.

Companies still keep senior people because they have context, judgment, and the ability to design systems. But junior jobs, internships, entry-level analyst work, content assistant roles, legal assistant work, research assistant roles, and support level one are easier to absorb first. This is bigger than a few missing job postings.

Entry-level jobs are the ladder young people use to learn real work. If the ladder disappears, Gen Z and Gen Alpha do not only lose a first job. They lose the environment where experience is formed.

Companies also become less willing to pay training debt. Why spend twelve to eighteen months training someone when AI can produce work that looks like a two or three year employee in many tasks? Experience then becomes a privilege.

People who arrived earlier have the context to orchestrate AI. People arriving later cannot enter the room to earn that context. The labor market starts looking like a fortress: senior workers and machines inside, young workers outside being asked to have experience before anyone lets them gain it.

The third truth: the AI boom is not equally for everyone. The media likes saying technology is being democratized. At the interface layer, that is partly true.

Anyone can open a chatbot. But at the power layer, the game is not that democratic. The real value flows toward whoever owns compute, data, distribution, models, chips, cloud, workflows, and channels.

The rest of the market can easily fall into two roles: paying users who rent intelligence every month, or cheap labor that supplies data, feedback, annotation, content, and behavior so the system can keep learning. I am not saying small players have no chance. The chance is not “I also use AI.” Using AI is now just the survival condition.

The edge is whether you have proprietary data, your own distribution, your own workflow, deeper customer understanding than the model, and the discipline to turn AI into an operating loop. Without that, you are renting the infrastructure of digital landlords while telling yourself you are independent.

The fourth truth: the “one person company plus AI” model may not be as beautiful as the pitch deck version. I still believe small teams have a speed advantage. But if everyone has the same model, the same templates, the same prompts, the same landing page tools, the same AI coding assistant, the same AI design assistant, and the same AI marketing stack, the barrier to entry moves close to zero.

When the barrier is zero, the market saturates fast. Products look the same. Content looks the same.

Small SaaS products look the same. Small agencies look the same. Newsletters look the same.

Courses look the same. At that point, “one person plus AI” is not a moat. It is just the minimum setup.

To survive, one person still needs what AI does not automatically create: distribution, trust, taste, insight from the real market, access to a real community, sales ability, the ability to choose a problem worth solving, and operational discipline. Without those things, a one person company becomes a faster machine for producing commodities. Speed does not save you if the market does not need another copy.

The fifth truth is the darkest one: the lower layer of society may get trapped in work that is cheaper, faster, and more tightly managed by algorithms. The phrase “14 million people doing three-yuan tasks” needs to be checked market by market and source by source, but it describes a very real feeling inside the gig economy. Humans become the physical substrate of the digital system.

Delivery workers race against routing algorithms. Data labelers complete tiny tasks. Freelancers fix small errors.

Support workers handle whatever the machine has not learned to handle. When AI takes over high-level cognitive tasks and robots are not yet cheap enough to replace physical tasks, humans can get pushed into low-price work that the system has not automated only because automation is not yet economically worth it. That is the ugly paradox: the more we talk about superintelligence, the more people risk living like batteries for the infrastructure.

The ethical question is not whether AI is intelligent. The question is who receives the productivity gain, and who gets turned into an optimization variable inside someone else’s dashboard.

By 2028, I do not think the question will be “will AI change the world?” It already has. The question is which layer of the stack you are standing on. If you are standing in the layer where you only sell time, repeat tasks, hold no data, own no distribution, exercise little judgment, and have no real relationship with customers, the risk is high.

If you can design workflows, control output quality, build distribution, understand human behavior, preserve trust, and turn AI into an operating system for real work, the opportunity is still large. The future does not belong to people trying to race machines on knowledge. Machines will win in many zones.

It also does not belong to people who deny AI and comfort themselves with “humans are still special.” Special is not a default right. Special has to be proven through real value. If AI can do most of what we do better, faster, and cheaper, we have to answer a hard question: what is the one value we are willing to protect, train, and bring to market so we do not become redundant inside our own society?