As a dad, I’m rethinking “learn to code” for my 12-year-old

Coding still matters, but the bigger question is what foundations children need first in an AI-shaped world.

Article link: https://open.substack.com/pub/johanosteyn/p/as-a-dad-im-rethinking-learn-to-code?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Audio summary: https://youtu.be/6EiZm3OwIco

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

I learnt to code when I was young. It was empowering. It taught me how machines “think”, how to break problems into steps, and how to build something real from an idea. But as the father of a 12-year-old, I’ve found myself hesitating before repeating the classic advice: learn to code. With AI now able to generate working code from a paragraph of instructions, the temptation is to assume coding will soon be obsolete, or at least no longer a reliable career path.

And yet, the more I read and watch, the more I see the opposite: not that coding is dying, but that the definition of what matters is changing. The question is no longer whether my son should learn syntax early. It’s what he should learn first, so that coding becomes a tool he can wield wisely, not a trick he performs mechanically.

CONTEXT AND BACKGROUND

For years, we’ve been sold two competing stories. One says software is the language of the future, so every child should learn to code. The other says AI will write most of the code, so learning is pointless. Both are too simple, and both can mislead parents.

The reality is that the software world is being reshaped, but not in a neat, linear way. Reuters captured one aspect of this shift in August 2025, describing how AI is already disrupting the “bootcamp to six-figure job” pathway that many people relied on, and how entry-level routes into development are tightening.

At the same time, we’re seeing new forms of coding literacy emerge. In South Africa, a practical, child-friendly approach to coding education matters even more because it intersects with access, opportunity, and confidence. A recent TimesLIVE story about a South African-developed learn-to-code programme gaining global recognition is a good reminder that “coding” can be taught as problem-solving, not just typing.

INSIGHT AND ANALYSIS

Here’s what I think many parents (and executives) miss: AI makes it easier to produce code, but it does not automatically make it easier to produce good software. Software is not just code. It includes requirements, trade-offs, security, testing, maintainability, and accountability. In real organisations, the hardest part is seldom writing a function. It’s understanding what the business actually needs, anticipating edge cases, and building something that doesn’t collapse under pressure.

That’s why the most honest “AI coding” commentary isn’t triumphalist. Business Insider recently quoted Andrej Karpathy describing a “phase shift” where he’s “mostly programming in English” using modern AI tools, while also admitting his manual skills can atrophy. The point isn’t to panic. It’s to recognise that ability will drift towards those who can direct, verify, and refine AI outputs, rather than merely accept them.

And the verification part is not theoretical. AI-generated code has security consequences. ITPro reported in October 2025 that one in five security leaders surveyed said they had suffered major incidents linked to AI-generated code. In other words, “the AI wrote it” is not a defence; it can be a liability.

IMPLICATIONS

So what should a 12-year-old learn first? My answer is “coding plus”. Before we rush a child into languages and frameworks, we should build foundations that survive every hype cycle: maths for precision, writing for clarity, and critical thinking for judgement. Add curiosity, patience, and the confidence to experiment and fail safely.

Then introduce coding as a way to express logic and build small things: games, simple apps, automations, robotics, and fun projects that make cause-and-effect visible. Only after that does AI become valuable, not as a crutch, but as a tutor and co-pilot.

This also means we need to be honest about risk. AI-assisted development tools can be powerful, but they expand the attack surface. Tom’s Hardware reported in December 2025 on serious vulnerabilities found across AI development tools, showing how autonomy can create new security pathways if not designed and governed properly.

CLOSING TAKEAWAY

I don’t think the question is whether my son should learn to code. The better question is what kind of thinker he should become. In an AI world, the advantage won’t belong to the child who memorises syntax fastest. It will belong to the child who can reason clearly, communicate well, test assumptions, and take responsibility for outcomes. Coding still has a future, but it’s increasingly a foundation beneath bigger skills: systems thinking, security awareness, and good judgment. If we teach children those first, they can learn any language later, and collaborate with whatever machines come next without surrendering their agency.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net