The day before Winter Holiday started, the Term 3 School Report came in. The moment I saw it, I was furious. Standing in minus-10-degree wind, I chain-smoked three cigarettes. Back home, I handed the report to my wife. She glanced through it and said: “Looks fine.”

Fine?

If you only look at the report itself, nothing screams disaster – the comments are “proper,” the language is “polite.” But that’s exactly what’s dangerous in the AI era: the more polished the surface information, the easier it is to stop thinking.

So I pulled out last year’s final-term School Report and compared the two. The problems became glaring: Math and Chinese had dropped from Significant Above. Other subjects had slipped too.

That evening I called my son into the study and gave him a serious talk. Teary-eyed, he sat there. I asked: do you know what the problem is?

He wiped his tears and said timidly: Math and Chinese?

That answer made my heart sink – he saw the “result” but missed the “structure.”

You have 30 minutes. Write down your current problems and solutions, like a report.

Half an hour later, he handed me a sheet of paper: one line per subject, no bullet points, no structure, no logic, all run together.

I cleared my throat:

Two big problems here.

  • Story Telling: Writing isn’t about filling space. It’s about expressing ideas clearly. How is anyone supposed to read this wall of text?
  • Insight: You’re looking at the problem too narrowly. Let me tell you what your real problems are.

What’s the Core Competitive Edge for Students in the AI Era?

Traditional education identified “core competencies” long ago: memorization, test-taking, producing standard answers, following procedures…

The problem: AI is crushing all of these at exponential speed.

When “solving problems” becomes a button, “writing essays” becomes a template, and “doing research” becomes a search box – grinding harder on these things is just competing with GPUs on endurance.

Strip away the competencies that will foreseeably be replaced. What’s left, in my view, comes down to three things:

Story Telling: The Top Productive Force of the AI Era

I put story telling first not because “liberal arts matter” but because it’s fundamentally a composite skill:

  • Language ability (clear expression)
  • Communication ability (making others understand and want to listen)
  • Social ability (understanding people and relationships)
  • Influence and leadership (getting people to follow you)

More critically, in the AI era, story telling is an accelerator for prompt engineering.

Same model, same tools – the gap often isn’t about “understanding AI,” but about whether you can articulate needs clearly, specify goals precisely, state constraints fully, and provide reusable context.

If you can’t tell stories, your prompts will always be: “Help me write…”

If you can, your prompts read like a good director’s brief: character, motivation, conflict, pacing, style, boundary conditions – sentence by sentence, steering the model exactly where it needs to go.

Math: The Foundation of Computation and Abstraction

Math isn’t about test scores. It’s about training a capability: abstraction, modeling, reasoning, verification.

AI can do the calculations, but it can’t replace your judgment on “should we calculate,” “what’s the right approach,” and “can we trust the result.”

The most valuable people in the future won’t be “those who can use tools” but “those who know what problem the tool is solving.”

Computing / Programming: The Handle for Wielding Tools

AI makes tools more powerful, but the threshold becomes deceptively strange: it looks like anyone can write code, yet nobody can build a system.

You can use AI to generate a script, but without a computing foundation, you can’t:

  • Decompose problems
  • Organize data
  • Understand boundaries
  • Debug errors
  • Build maintainable structure

You end up stuck at “copy-paste and pray it runs.”

Understanding the Current Problems

I told my son something blunt that day:

Significant Above means it’s your competitive edge. And now, it’s gone.

Many kids (and parents) treat the Report as a “scorecard.”

But in the AI era, a Report is more like a Competitiveness Dashboard: what matters isn’t any single score, but the trend of your capability curve.

The Vanishing Competitive Edge

His answer “Math and Chinese” showed he only looked at the red text and bold font.

The truly alarming thing: what used to be a significant lead has suddenly become average.

Two possibilities:

  • The prior lead was “coasting” (luck, foundation, teacher quality, test format)
  • After entering a new phase, the learning method didn’t level up, and peers started catching up

There’s no “coasting” in the AI era. AI flattens basic-skill gaps fast. What remains is a battle of learning systems.

The Feedback Loop Is Broken

Teacher comments keep mentioning “attitude,” but I’d translate that as: the feedback loop is broken.

  • No goals (why am I doing this)
  • No feedback (how am I doing)
  • No correction (what’s wrong, how to fix it)
  • No retrospective (how to do better next time)

So the kid defaults to the lowest-effort strategy – faking it. Yelling won’t fix this. It requires systematic rebuilding.

Building the Feedback Loop

The Underwhelming AI Coding Class

After half a year of online coding classes, I figured he could write something by now. I was planning to walk him through LeetCode, teach some data structures and algorithms. So I gave him the simplest problem:

Find and print all even numbers in an array.

He froze. Had no idea where to start.

Barely containing my frustration, I asked:

What have you been learning in your weekly AI coding class?

He said: “We used Python for image recognition and voice recognition, but everything was inside the teacher’s software. I’ve never written code in an editor (PyCharm).”

My reaction: we’re doomed.

Writing “fill-in-the-blank code” in a sandboxed environment doesn’t teach programming. It teaches button-clicking. That’s like learning to drive in a self-driving car for six months without touching the steering wheel.

So I told my wife: no more coding classes. I’m teaching him myself.

I started preparing lessons whenever I had time, and eventually put together an online deck: https://cs.johnsonlee.io

Starting from zero, teaching real programming in a more engineering-oriented way to solve real-world problems.

I’ve watched plenty of kids’ coding videos online. Most start with Scratch or games. From where I stand: somewhat useful, but only somewhat.

Coding is the path you have to walk. Rather than spending tons of time learning Scratch’s “block syntax,” it’s better to start engaging with how the real world expresses things.

A Diary Entry That Nearly Did Me In

His Chinese teacher had been saying his writing needed improvement, so I planned to have him write a weekly journal – in Chinese or English, either was fine.

Last weekend we went skiing, and he wrote a diary entry. The moment I read it, I nearly coughed up blood – pure play-by-play:

Had breakfast, had lunch, did stuff in the afternoon, went home at night, the end.

I said to him:

Son, “diary” may literally mean a daily record, but do you think anyone wants to read a blow-by-blow account?
Think in reverse: what kind of content would people actually enjoy reading?
If it’s an exam, what makes a high-scoring essay? What are the criteria?

I twisted the knife:

Everyone eats breakfast. If what you write is as common as air, it gets ignored.
When you read other people’s work, what’s more interesting?
Something novel, something different from what everyone else writes – different experiences, different thoughts… Writing is storytelling.

So he rewrote it following my guidance. You think that’s the end? The show was just getting started.

I had him open ChatGPT, paste both versions of the diary, and ask it to compare and critique them. When he saw the new version receive a much higher rating, he broke into a proud grin.

One of the things AI does best is serve as a child’s instant review panel. But the prerequisite: the child has to learn to write the “story” first.

Mentor vs. Classroom Teacher

Though I teach my son coding, my real role is more like:

  • Observing and identifying current problems – what’s primary, what’s secondary
  • Analyzing problems, identifying core competitive edges – what’s the actual winning factor
  • Building the feedback loop – how to keep getting stronger
  • Using AI to establish evaluation standards – what counts as good, what doesn’t
  • Constant supervision, fighting human nature: laziness, avoidance, procrastination, shortcuts

Closing Thoughts

After that study-room talk, I took a step back and thought about it myself:

What I want isn’t for him to “bring his grades up this time.” I want him to spend the next three years building a set of capabilities that won’t become obsolete.

But the hardest part of passing on everything you know isn’t the method – it’s the pacing.

A child isn’t your project. He won’t ship on schedule just because you wrote the spec.

The reality is: you teach three times, he absorbs once. You reason with him, he reacts emotionally first. The more you push, the more he retreats.

So I set three ground rules for myself:

  • Incremental progress: tackle one key point at a time, no grand overhauls
  • Let him participate in setting the rules: he’ll only maintain a system he helped design
  • Use AI for instant feedback, not as a substitute for thinking: AI is the mirror, not the crutch

I’d love for him to get it sooner rather than later. But I also know that real growth is never a sprint – it’s a marathon.

After all, the prerequisite for “passing on everything you’ve learned” to your child is learning, first, not to treat him as an extension of yourself when he stumbles. And that might just be the first lesson of being a mentor in the age of AI.