The Giraffe You Can't Describe

A friend asked me a question over dinner in Taipei last week that I've been turning over ever since.

"Given how fast AI is moving," he said, "training kids to be traditional knowledge workers feels like a dead end. If a family has the money to absorb some trial and error, should they just skip the system entirely? Let the kid jump straight into human-machine collaboration? They're AI natives anyway."

I get the impulse. I even agree with the direction. But the conclusion—that you can skip the foundation and just start orchestrating the machines—is dangerously wrong. And I say that as someone who makes his living orchestrating machines.

The Tool Is Dead, But the Scar Tissue Isn't

In the traditional industrial era, one visionary leader commanded a team of human tools to execute a task. In the AI era, that same leader commands an army of agents. The human tool—the person who just knows the software, just follows the procedure, just has the domain knowledge—has been squeezed out of the ecosystem.

If you train yourself, or your child, to be the perfect corporate tool, you will waste years and capital only to discover you're obsolete before you ever got started.

The three sacred pillars of the traditional professional:

• I possess deeper domain knowledge.
• I am a master of specific software tools.
• I am intimately familiar with standard operating procedures.

In the future, these three pillars are practically worthless. AI has shattered the barriers of knowledge, tool mastery, and procedural execution.

Theoretically, you no longer need to write code yourself. You no longer need to grind through junior, mid-level, and senior developer roles. You can jump straight to Systems Architect or Product Manager, orchestrating your AI army.

But note my phrasing: in theory.

The Fatal Flaw

Here's the gap in that logic: If you've never actually done the foundational work, where exactly does your architectural judgment come from?

How can you possibly be a competent Product Manager if you've never felt the pain of a failed deploy at 3 AM? How can you architect a trading system if you've never watched an order book move in real time, felt your stomach drop when the spread widened, and learned to recognize the pattern before the algorithm did?

Great generals rise from the infantry. Great prime ministers rise from local governance. Do you think military geniuses like Xiang Yu or Huo Qubing were born with tactical knowledge? Judgment isn't innate. Judgment is born from scars, repetition, and intimate familiarity with the work itself.

The Giraffe Dilemma

Imagine a person born completely blind. Can you use words to make them truly comprehend the physical world? You describe a giraffe—long neck, spotted coat, four legs, eats from trees. Can they actually picture it? Can they feel the absurdity of its proportions, the specific way it moves?

Now take a person born with sight who lost it in adulthood. Describe a giraffe to them, and their brain immediately recalls the exact image. The experience is already encoded.

I use this analogy to explain why foundational experience is non-negotiable.

Take my field: high-frequency quantitative trading. Even before the current AI boom, humans weren't manually clicking buttons to execute trades. Algorithms opened and closed positions automatically.

So under these conditions, is it still necessary to stare at a live trading terminal?

For me? No. Not practically.

But I have spent over 10,000 hours staring at order books—watching Bid 1 through Bid 20 and Ask 1 through Ask 20 constantly shifting, learning to read the micro-structure of liquidity, feeling the rhythm of the market. Because of that, even when an algorithm executes the trades, I know exactly what's happening under the hood. I might miss a microsecond detail, but my foundational judgment remains intact.

Now take a brilliant math student who's never looked at a live trading terminal. His intellect is flawless. He understands every quantitative algorithm on paper.

What is he missing? He's missing the giraffe.

He's never felt the visceral reality of a fluctuating market. Can I discuss trading algorithms with him? Absolutely. But if I say, "I have to step out for an hour. The algorithms are running. You are in charge," what happens?

He panics. He's terrified. His terror comes from a profound lack of foundational experience. He knows the theory. He doesn't know the animal.

The An Lushan Test

During the Tang Dynasty, the powerful warlord An Lushan was kept completely in check by the brilliant Prime Minister Li Linfu. Why? Because Li Linfu had vast, brutal, practical experience. An Lushan later admitted that whenever he spoke with Li Linfu, he felt completely exposed—as if the Prime Minister could see right through him. He didn't dare rebel.

But when Li Linfu died and was replaced by Yang Guozhong, everything fell apart. Yang Guozhong had all the theoretical management titles, but his inner core was hollow. He was a man trying to control a beast he had never actually wrestled. An Lushan immediately sensed the weakness and rebelled.

AI is our An Lushan.

Today, you might feel like you can control AI without knowing the foundational skills, because the old guard—the Li Linfus—are still around designing the guardrails. But what happens when the Yang Guozhongs are left alone in the room with the machine?

Can you really claim that foundational experience is meaningless?

I force my youngest engineers to stare at live trading terminals. Not because it has any practical use—they will never execute a manual trade in their careers. I make them do it because their entire professional life will be spent managing algorithms. They need that missing page in their cognitive history. They need to have seen the giraffe.

We don't calculate complex math by hand anymore; we use calculators. But we still force children to learn basic arithmetic. Why? Because without that foundational struggle, you cannot trust the machine's output. You must know that if the autopilot fails, you can still land the plane manually.

The 90% Rule

I'm not rejecting traditional education or foundational training. It's a matter of degree.

In the past, you had to score a 99 out of 100 on traditional skills to prove you were a useful tool and get hired. Score a 90, and you were out.

Today, you only need to score a 90. You can take that remaining 9% of your time and energy and invest it entirely into human-machine collaboration—learning how to make strategic judgments, how to command an AI army, how to operate at the layer where the machine can't.

This is the correct path.

But it absolutely does not mean that because AI exists, you can score a 20. Or a zero.

If you score a zero on the foundations, why should the AI army take orders from you? Isn't it highly likely that your commands are actually detrimental to the system? If we remove you, the AI army will probably perform better.

The Mirror

Ultimately, human value is still determined by humans. Your market price is dictated by the fact that [You + AI] is more powerful than [Your Competitor + AI].

AI is just a mirror. It reflects the competence of Li Linfu, and it exposes the utter uselessness of Yang Guozhong.

Which one are you training your child to be?

— James, Mercury Technology Solutions, Taipei, May 2026