There were two distinct moments in my life when my parents talked me out of two careers. The first was when I was young—like, elementary school young. I wanted to be a Disney studio animator. Not in the vague, wistful way kids want to be astronauts, but genuinely. There was nothing I wanted more than to be part of the team that built entire new worlds from graphite sketches and an easel. My parents understood my passion, and they weren't unsupportive. They just looked at where technology was heading and made a calculated call: computers were coming for that world, and a kid with a pencil wasn't going to outrun them. Sure enough, Toy Story dropped, then later on Pixar became its own thing, and the case seemed to close itself. The second was computer programming. Nearly a decade later, same conversation, different noun. The landscape was moving at a rate that was unfathomable to someone who wasn't used to exponential growth. Meanwhile, the thought was that while I was still in high school, whatever programming language I learned would probably go stale in 4 years, and certainly obsolete by the time I entered the work force.
I was never resentful. Again, their logic seemed sound...if you granted them their underlying assumptions about how careers worked that were honestly understandable for their immigrant and generational vantage point. Besides, I was young and had neither the vocabulary nor life experience to push back on their objection, so what did I know really? But something about their framing felt off in a way I couldn't articulate yet.
After many more years (much of it spent in college and around specialized craftspeople), I finally figured out what was "off" about my parents' framing of what was and wasn't good for my career path. It was the structure of that framing; not so much the conclusion they came to. Namely, they were predicting the future of visible outputs (character rigging, coding syntax, etc.), but overlooked the more important aspect: developing computational thinking (which in a way trivializes those outputs). Those with unstable job prospects weren't necessarily losing to those who knew more recent programming languages or had better syntax; they were losing to those who could better leverage computational thinking to adapt to new situations. I'm sure the same goes for animators; the tools are just that—tools. Conversely, looking back on all the times I felt FOMO, it really came down to wishing I had the right mental models—the literacy—to speak intelligently with specialists around me, rather than wishing I could execute the same tasks.
Now that I have a kid of my own, this came back to me when one evening my wife asked me if I would want to enroll our son in after-school programming classes.
Looking back on that evening, I'm still not sure if she was asking that in light of the ongoing discourse around AI and many voices (who those voices are is important, but a bit outside the scope of this post) declaring that coding is dead as a prestigious skill set1. Regardless, it was a question that could invite an impulsive answer in either direction. Of course, fundamentals never die vs. Why bother when AI writes cleaner code than most junior developers? But I think the deeper question worth asking that she hinted at is, What does learning programming do to the mind that learns it? And from a more existential angle, Would that cognitive artifact survive the automation of the outputs it was originally trained to produce? What cognitive skills would be necessary to thrive in a digital world where AI blurs the lines between creator and products of creation, and even if I can't predict what those skills will be 20 years from now, how can I at least set our son up for success today? Come to think of it, I myself will still be working 20 years from now!
Subtly modifying her question in my mind, I replied to my wife, yes, I'd want our kid to learn some coding.
Regardless of where my wife's question was coming from, to me, the value of learning to code has never been the code. At its core, it's the confrontation with abstraction. You have to hold a complex system in your head, decompose it, and then reconcile your mental model of it with what it actually does when you run it. It's a discipline to hone the habit of asking: what is this system actually doing vs. what do I believe it's doing? Debugging isn't really a technical activity; it's a discipline in figuring out where your thinking broke down, and that habit of mind doesn't go away when AI writes cleaner boilerplate than a junior dev. The differentiating skill shifts from "can you write the solution" to "can you write down your thought process about the problem clearly enough that the solution becomes obvious."
These are not software engineering competencies. They are epistemological ones. Given that, I'd argue that these kinds of systems-thinking competencies are (should be) transferable to any other discipline or problem space.
Okay, let's get more pragmatic… What might that mean in terms of economic opportunity?
I forgot which talk I found this in on YouTube, but Andrej Karpathy alluded to the Jevons Paradox manifesting in the AI space. That is, when a technology makes something cheaper or easier to consume, demand for it tends to go up rather than down. Well, isn't that...obvious? I guess whether that's obvious or not comes down to your assumptions about whether efficiency inherently multiplies opportunity (we do tend to take that for granted these days, in a highly connected, capitalistic economy). In the classical school of thought, the expectation is that if a (steam) engine requires 50% less coal to produce the same amount of energy, individual consumption drops by 50%. But Jevons pointed out that paradoxically, steam engines made coal so useful that everyone wanted more of it—well, "of course", right?
The example Karpathy used in his talk was how the proliferation of ATMs initially led to a fear of bank tellers becoming obsolete, but actually ended up significantly reducing the overhead of bank branches, thus leading to more bank branches being built and thus increasing demand for tellers. It seems to me that the paradox is even less obvious in AI; the discourse about coders being replaced by AI essentially asserts the 19th century logic that if an artificially intelligent engine requires 50% less manual code to produce the same amount of software, the deployment of manual code drops by 50%. Yet counter to the collective anxiety about coder obsolescence, the price elasticity of software is such that, apparently, the job market is actually seeing a demand increase for programmers.
In other words, the automation of coding may not shrink the universe of people who need to think computationally. There already seems to be plenty of evidence today that AI is actually expanding it.
That begs the question, who exactly are the voices declaring coders obsolete in the first place?
That might be a conversation for another time, but it's worth briefly mentioning here a take from Jensen Huang where he said that being smart (however you might want to define that, but let's for sake of argument align that with being an exceptional programmer) is a commodity. If AI is solving software programming, then the long-term "delta in smartness" between man and machine is, to quote Jensen, "someone who sits at the intersection of being technically astute but human empathy, and having the ability to infer the unspoken; the around the corners; the unknowables...to be able to preempt problems before they show up just because you feel the vibe. And the vibe came from a combination of data analysis, first principles, life experience, wisdom, sensing other people."
It's not the same as computational thinking, but I think there's massive overlap that comes down to being well-practiced in making non-obvious connections out of a huge pool of disparate information. And I'm inclined to say that systems thinking, though probably not a strict prerequisite, significantly enables that.
So yeah, I would very likely enroll my son in some kind of educational program that immerses him in that kind of thinking, whether it involves coding or not.
And honestly, I'd want him to develop the same hunch I had as a kid, that stubborn, formless instinct that the person behind the tool is never fully replaceable, but with more vocabulary than I had to defend it when someone tells him his skills are about to be obsolete. Wanting that for him is one thing, but arguably the harder part is just being honest with ourselves that we're choosing which lenses to hand our kids while operating with incomplete information, in real time, the same way every generation before us has had to.
Maybe that's just parenting. Huh.