Nov. 5, 2026
Previously, I argued that everyone should learn to code. (Well, everyone who wants to learn how to think better, anyway.)
I should point out that many, many people disagree strongly…
I Googled this, and I found loads of top hits where the vast majority of individuals argued that No, coding is NOT essential.
- In the past, many people said things like, “No, because there’s lots of work that doesn’t involve coding.”
- More recently, people say things like, “No, because AI can do that for us now.”
- I saw one comment along the lines of, “If we do all need to code, then computers have failed as a concept.”
- When asked about the benefits of coding, one person even said there were “absolutely none.”
- Someone said, “%$#! no.”
That’s fine, but I think they’re all wrong.
The crux of their confusion (in my view) is in thinking that the value of coding is limited to employment skills.
In fairness, it makes sense that people would assume that’s all there is to it, because in the past it was very useful for that specific purpose. (Arguably it still is, though I don’t know enough to say.)
One particularly common response to the claim, “Everyone should learn to code,” is to list any number of other activities (e.g., building cars, forging metal, painting, writing novels…), which obviously not everyone “ought” to learn to do.
To be clear, if we’re talking about computer programming more generally, as in being able to design entire software applications…then I totally agree that not everyone needs to learn that.
But if by coding we simply mean being able to express ideas in some programming language…then I completely disagree.
Consider it like this…
I grant that you don’t need to know how to write a novel, but you really do need to know how to read and write, at least at a basic level, in order to function optimally in society.
In the same way, you don’t need to know how to program an entire software application, but I’d argue that you do need to know how to express concepts in digital form in order to think optimally.
As I’ve said before, it’s like eating your vegetables: It’s not a moral imperative, but it’s good for you.
Among those who do (or, prior to the increased popularity of LLMs, did) believe coding was universally beneficial, the reasoning usually relates to either getting a job or improving logic and problem-solving skills. Additionally, coding is often promoted as a way to understand computers better.
I don’t disagree with these points, but I still think they miss something more fundamental…
In my view, the reason we should all learn to code is so that we can better clarify our understanding of the world.
It’s not enough to be able to express our ideas verbally. Our verbal explanations often gloss over conceptual gaps without our even realizing it.
After saying “LLMs are predictive text on steroids” for the millionth time, we start to forget what a dramatic oversimplification that statement is.
And we slide into the trap of thinking that, because the words in our explanations come so easily, therefore we must understand deeply.
“But how does predictive text work in the first place?”
“And what is the mechanism that steroids represent in this metaphor?”
Either of the above questions would be crushing for most people who liken LLMs to predictive text.
Our underlying mental models are not as shored up as we like to think they are. This is one of the primary reasons that coding is so beneficial.
Natural human language allows us to mask gaps in our understanding, because words are simply labels for concepts. You can utter a word regardless of whether you actually grasp its meaning with any depth. And entire statements pass our lips effortlessly, with no inherent check for either logical consistency or even the speaker’s own comprehension.
Nor is there any guarantee that a listener will point out logical inconsistencies. Often, if it “sounds right,” we assume it is right—and that goes for speaker and listener alike.
In contrast, when we force ourselves to articulate our thoughts in a language consisting of nothing but explicit unambiguous instructions, the conceptual gaps leap out.
In a world where surface-level understanding is increasingly the domain of AI, we’re called upon to understand more deeply, not less.
Coding offers a highly flexible medium for putting that to the test.
