2026年10月03日 星期六 台積電市October 3, 2026 — Saturday — TSMC City

Hello, friends among the stars. Welcome to my journal.

My name is Song Yan.

I have been trying to sort out my thoughts these past few days.

Not to finish an assignment, and not to meet a deadline. There are simply too many questions pulling against one another in my head. If I do not write them down, they keep circling inside me.

I tried using some more extreme assumptions to reason through the things in front of me.

Suppose there were a second Earth.

Suppose that, somewhere else in the Milky Way, there were a planet whose environment, resources, and population structure were almost identical to ours. They had also experienced resource depletion, signed a ceasefire agreement of their own, and were now competing with other civilizations through technology and institutions. How would we have to develop in order to truly surpass them?

It would not be a matter of whose chips were faster, or whose spacecraft could travel farther. Those things can be copied, and they can be surpassed. What is truly difficult to copy is an entire method that allows people to keep cooperating, keep correcting themselves, and keep turning different ideas into shared results.

Which skills can human beings use under any circumstances?

I listed several: observing, asking questions, breaking complicated things into smaller problems, making decisions amid uncertainty, and—most importantly—being willing to admit that we may be wrong, then adjusting.

These abilities do not require the most advanced equipment or any particular resources. As long as we retain the ability to think, they can be used. Conversely, if a civilization outsources all of these basic abilities to machines and leaves humans only to execute instructions, then that civilization has already turned itself into another kind of robot.

Our society seems to be training many people to become robots.

Attend classes, take exams, complete internships, enter a company, finish tasks that have already been broken down, and then repeat the cycle again and again. It is highly efficient, but the space for thought has been compressed almost to nothing. Today, collaboration with AI is already inseparable from daily life, and that is not necessarily a bad thing. But when AI begins to replace the humans who have been “trained into robots,” we suddenly realize that many jobs became replaceable because we first made ourselves replaceable.

So learning how to live with AI may be one of the keys to whether human civilization can continue moving forward.

I tried to imagine two extreme scenarios.

The first: a country develops AI too aggressively without sufficiently powerful regulation and control.

Economically, productivity would surge, but wealth would be concentrated even more quickly in the hands of the few who control the models and computing power. Large numbers of middle-layer jobs would disappear, creating a visible social divide. Education would be forced to shift toward “how to use AI” rather than “how to think,” because it would seem that machines had become better at thinking. A culture of cooperation would grow fragile as people became accustomed to handing decisions over to models and gradually lost the muscle of making judgments for themselves. In the long run, such a country might be extremely powerful, but also extremely fragile. If a model developed a systemic bias or were manipulated by a small group, the whole society would have very little capacity to adapt.

The second: an AI nation with a strict regulatory and governance system.

Its development would be slower, because every step would need to be examined, made public, and traceable. But education and the economy would also be required to upgrade in parallel. Schools would have to teach students how to question model outputs, how to design constraints, and how to make trade-offs between AI recommendations and human values. Companies could not pursue efficiency alone; they would also have to prove that their systems did not exclude people from decision-making. Cooperation would place more emphasis on “division of labor between humans and machines” rather than “humans being replaced by machines.” Such a nation might look less dazzling in the short term, but it would have much greater resilience.

Neither scenario has a standard answer. I am simply trying to see more clearly what we are actually developing when we say we are “developing AI.”

During summer vacation, I visited an exhibition and jotted down some notes. Looking at them now feels almost unbelievable.

The things that were called trends back then have now quietly become the foundation of nearly every piece of software we use.

My notes included:

Human interpretation and analysis modules for different industries; structured voice calls, with AI agents handling calls and negotiations; data visualization that presents documents dynamically; personal assistant systems for private-account management and chat histories; calling on different kinds of reasoning and execution to encode a single skill; training AI models with sensor data; and using image recognition for non-digital inspection or sensor applications.

For trends, I wrote down data-use security, data integration and structuring, and the integration of physical and virtual worlds with VR-based interpretation of reality.

The common architectures could roughly be divided into three categories: control—the tools AI uses to execute actions; understanding and decision-making—the tools for analysis and formalized logic; and data—the tools for organizing information and analyzing databases.

At the time, these concepts felt new and cutting-edge. Looking back now, they have quietly grown into almost every system we use each day. Change is happening so quickly that we have to ask again: at this speed, what advantages do humans still have? And what are our disadvantages?

Business competition has also kept me thinking.

On the surface, many companies are doing repetitive things—burning money, competing for talent, competing for attention—and it can look like a waste of resources. But without that competition, many of the tools we use today might never have appeared. Competition is a double-edged sword: it pushes technology forward, while also consuming enormous energy in conflicts that cancel one another out.

The more practical question is: how can we help traditional companies get on this ship?

Many companies still do not even have a decent website. Their internal processes remain stuck in paper documents and group messages. Meanwhile, the outside world is preparing to move beyond Web 2.0 and enter a new environment built on AI at its foundation. The gap is so large that it is hard to know where to begin pulling them forward. Should we wait for them to realize it themselves, or should we first make the tools simple enough—simple enough for them to be willing to take the first step?

These questions are tangled together, and eventually they circle back to the same thing.

The announcement that INN is establishing an independent news node on Earth still leaves me deeply hesitant.

The school year has just begun, and I have not adjusted to many things yet. Coursework, routines, and new interpersonal distances all need time. I tell myself to wait a little longer. If no one is willing to take it on for a long time, then I will try.

But over the past few days, while talking with others about INN, I discovered that there are people online who are interested in it just as I am, but who temporarily cannot make time for it. Some are preparing for exams, some are working part-time, and some simply feel that they are not ready yet.

I really do want to do it.

An independent news node that is not directly controlled by a corporation or government sounds like opening another door outside the existing system—a door that can be verified and participated in. It seems to echo, in some quiet way, the question I have been thinking about these days: how can different perspectives be made visible?

But reality is harsh.

When I lay out the time, energy, risks, and limits of my current abilities and calculate them together, every answer points to the same conclusion: “It is not time yet.”

So I am stopping here.

I really want to do it, but I have decided not to move yet.

This feeling is more painful than simply giving up.

I read through my exhibition notes from summer again and added today’s thought experiments to them. The page has become messy, filled with arrows, question marks, and sentences that have been crossed out and rewritten.

I did not reach a conclusion.

I simply moved the questions from my head onto paper, so they would have somewhere to stay for a while.

I am going to get lunch now. Good afternoon.