I thought sending my kids down the international school route would let them escape the hyper-competition back in China. But after moving to Seoul, I realized the rat race here is just as intense. Despite the environment, we’ve been deliberate about not over-scheduling – we only sign them up for classes they actually ask for. After talking with other parents at the international school, I noticed their logic was almost identical: replicate the last generation’s playbook for international students – stack extracurriculars, pile up credentials, and treat certificates as the keys to elite university admissions.

But having witnessed everything that’s happened over the past three years, I’m starting to think this playbook may not survive the times.

From OpenAI’s emergence to the release of Gemini 3.0, AI’s iteration speed doesn’t look like technological progress – it looks like a rewrite of society’s operating system. Facing this kind of acceleration, I can’t help asking myself:

Will the world ten years from now still need the kind of “international kid” we’re so busy cultivating today?

The answer is not encouraging.

The Next Decade: How Will Demand for Talent Change?

A picture is forming in my mind – increasingly clear, increasingly cold:

Ten years from now, 80-90% of people in society will become “non-contributors.”

Not because of laziness or lack of credentials, but because the frontier of production will have expanded beyond what humans can keep up with.

AI is growing exponentially in processing information, understanding language, generating content, analyzing data, even writing code – while human improvement remains stuck at the faint, generational pace it has always been.

This means most future value creation will be done by a small number of people who command technology, not by a large mass of ordinary workers.

The remaining 80-90% will be forced to rely on social systems to maintain dignity – and may even be redefined as “structurally redundant population” under automation and welfare frameworks.

We are not experiencing this kind of shift for the first time, but this is the first time the gap between humans and machines is widening at an unbearable speed.

Put differently:

Society will no longer need “large numbers” of talent – it will only need “a tiny minority who can truly wield technology.”

This structural change will hit the current education system head-on. The extracurricular classes, competition certificates, overseas experience, and even diplomas that once served as “keys to the door” may open nothing important a decade from now.

Because what society wants is not “well-roundedness” or “diversity” – it’s the rare few who can make AI work for them.

Non-Contributors Will Keep Growing

If we look back at the previous generation’s education system:

  • We poured enormous effort into training “people who can execute”
  • We used standardized tests to filter for “people who can memorize”
  • We rewarded “people who can repeat”

These are precisely the capabilities AI is best at – and where humans have no chance of winning.

AI doesn’t tire. Models don’t forget. Compute doesn’t complain. In every area we thought “required hard work,” it crushes us effortlessly: copywriting, information synthesis, research, code drafts, project plans, logical analysis, even strategy generation.

The result: work that used to take 10 people now takes one person plus a few models and robots.

Society doesn’t need that many people anymore.

But the education system is still mass-producing “replaceable people.”

That is the structural tragedy.

Knowledge Learning No Longer Depends on Teachers, but on Mentors

Knowledge is becoming as accessible as air. A child asks AI a question and gets an explanation potentially clearer than any textbook, complete with example problems, step-by-step walkthroughs, and extended discussions.

The scarcity of knowledge has vanished, teachers’ authority has declined, and the logic of learning is being redefined:

  • Before, teachers told you “what to learn.”
  • In the future, AI will tell you “how to learn.”

So what is a mentor’s role?

A mentor helps the child:

  1. Build capabilities that won’t be replaced long-term
  2. Establish an effective feedback loop to continuously track competitiveness

Many people confuse “mentor” with “tutor,” but these are entirely different things.

  • A teacher solves homework problems; a mentor solves life-direction problems.
  • A teacher transmits knowledge; a mentor transmits judgment.
  • A teacher answers questions; a mentor helps you ask better questions.

In an age where AI knows more than any teacher, what’s truly scarce is someone who can tell you “what you don’t need to learn.” If the direction is wrong, effort is meaningless.

Society No Longer Needs Masses of Ordinary Graduates

Many industries are undergoing a quiet revolution, and the direction is remarkably consistent – AI is shrinking “teams of dozens” into “a few people plus a model.”

This isn’t the future. It’s happening now.

One-Person Teams

I recently saw someone on X share their experience:

Just hired four new employees

Gemini handles the frontend
Claude Code handles the backend
Codex handles architecture
Lightning Talk handles typing for the AIs to read

Almost hitting my management capacity limit

It looks a bit comical, but it is genuinely happening.

Biotech: AI Has Turned “Running Experiments” into a Computable Problem

Drug development used to be an extremely long and expensive path:

  • Run experiments, fail, run more experiments, fail again.
  • Every step required PhDs, postdocs, and researchers investing enormous amounts of time.

But now more and more biotech companies are using AI for “virtual experiments”:

  • Feed the model millions of molecular structures
  • Let AI predict which ones have drug potential
  • Send the shortlisted candidates to the lab for validation

This transforms R&D from “try once, wait once” to “compute once, screen millions.”

For example, one of this year’s Nobel laureates – David Baker’s team at the University of Washington – recently used RFdiffusion, a generative AI, to design new antibodies from scratch. A process that traditionally required animal immunization, random screening, and months or even years of optimization can now potentially be completed in weeks. Traditional roles in specialized bio/chem labs may no longer need as many people – one or two people plus AI could be enough.

Similarly, in the field of protein structure prediction, AlphaFold solved the decades-old folding problem almost overnight. Many research institutions have since reduced headcount in basic structural research and redirected resources toward AI-driven design and synthesis.

This trend signals something:

What will truly be valuable in the future is not “how many experiments you can run / how many reports you can write / how much data you can organize,” but “whether you can wield AI as a tool and creative partner.”
The conventional “diligence + diploma” combination is no longer the key. What’s truly scarce is “people who can create, design, and command systems.”

What Is the Way Out for Future Education?

I believe only two directions are genuinely effective.

Find the Right Mentor: Direction Matters Ten Thousand Times More Than Effort

Children of the future won’t win by working harder than others. They’ll win by finding the right path sooner.

A good mentor can help a child:

  • Understand the capabilities the AI age truly demands
  • Build long-term knowledge structures rather than grinding skills
  • Avoid investing in competition tracks from the old era
  • Develop their own technological edge
  • Become one of the “irreplaceable people” ten years from now

The era has changed, but direction always matters more than speed.

The key to the future is not a resume – it’s irreplaceability.

And a mentor is the key to building that irreplaceability.

Consistent Physical Exercise: The Body Is the Most Undervalued Competitive Advantage

As AI pushes skill-level competition to its limits, the real differentiators among humans will be:

  • Focus
  • Stress tolerance
  • Willpower
  • Mental resilience
  • Capacity for high-intensity learning

These foundational capabilities are not trained in cram schools – they are built through sustained physical training.

The body is not just a container; it’s the operating system. The stronger the body, the faster the upgrades; the weaker the body, the more unstable the system.

In an era of rapid change, this is the key to survival.

Only someone with a resilient body has the right to keep up with the pace of the AI age.