What the hell do you mean?

Good question! My answer will certainly feel... broken compared to the usual "good code is when you follow DRY!" or "good code is when you create 50 quintillion classes and make everything OOP!" or even... the devil's advice itself: "good code is when EVERYTHING is serverless, and EVERYTHING is Javascript (or at least some form of web technology)" (ahem ahem, electron).

Most people these days learn programming via YouTube, or God forbid, AI, which... just isn't right. Recently, programming has essentially become "ship X in Y time, or else you are a bad developer", which just isn't how programming should be done. Of course, if you have a job or something similar, then that is how you should think about it (unless you really enjoy being homeless). However, if you are just doing it as a hobby then it is 100% the wrong way to approach it.

Cool... why is it a problem?

Well... it isn't. How we learn things constantly changes. When the printing press was invented, suddenly people started spreading and learning knowledge via written texts instead of "I heard about this from a guy who knew a guy who knew a dragon who told this story". Similarly, with the explosion of the internet, books became less popular.

However, what I'm criticizing isn't the way people learn programming but how they learn it. Basically, it isn't about the medium; it's instead about the depth (I swear I'm not ChatGPT).

My personal opinion is that as we go on, we have started to value thinking for ourselves less and less (this applies outside of programming but that's for another time...) and this has made it so we value abstraction, memorization, and of course, our most loathed and despised: prompting AI.

Well... AI is the future isn't it?

No. Well, yes, but no. You see, in the end, your brain is a muscle. If you don't train it, it will get weaker. The more AI you use, the dumber you get, whether you realize it or not, and more often than not, you will not realize it. We have seen this happen to younger generations in more "developed" Western countries, like the U.S.A.

The same happens with "tech literacy" as well. The more you outsource thinking to machines, the more your brain thinks, "hm... I haven't had to think about [x] in a long time... why keep those 'pathways' alive? let's... just forget that."[1], and boom, you grow 0.1% dumber.

Additionally, there is a reason I used '0.1%'. It's because it sounds so small and insignificant... because, well, it is. But that small insignificant number adds up slowly, till you are too far gone to recover and have no choice but to outsource to a machine.

Now, trust me, I'm not saying using AI makes you instantly dumber, and I'm not saying you should NEVER use AI, EVER. AI certainly has its uses. For example, even in my project: llmpp I used ChatGPT to generate the Makefile from my custom meta-build engine's syntax. Does that mean I have instantly forgotten how to write a Makefile? Unfortunately not.

But... something something printing press???

The printing press is very often used as an example to justify AI... and it's wrong. The difference between the printing press and AI is that one of these technologies doesn't think for you. You have the world's knowledge either way, whether you go to an AI or a library, but the difference is that AI spoon-feeds you all the knowledge you want.

When you read a book, you have to somehow synthesize that information into knowledge. You parse the sentences, you 'internalize' the logic, you think about the concepts and so on. With AI, on the other hand... there is none of that[2]. It skips the whole "processing" part of learning.

Think of it like a calculator and a GPS. With a calculator it helps you do math, yes, but you need to understand the logic behind the calculation. With a GPS, on the other hand, it just... tells you where to go. If you use only your GPS for a few years (or even a few months) without figuring things out on your own, you lose general spatial awareness. If you then suddenly stop using GPS, you will get lost more easily (Read More[3]).

Friction creates growth in our brains, and the less friction there is, the slower our brains get[4]. AI completely bypasses this friction. That is the problem. Read books.

Oh great... old man yells at cloud. How original.

LISTEN HERE YOUNG MAN- I mean, no. Here's the funny thing, I am still in high school. The only reason why I think I "deserve" to speak on this (at least a little bit) is because of the way I learned programming. Most of what I learned comes from books I borrowed from my family... out of which exactly zero are not older than me.

Hell, I have several books on my shelf about J2SE/Java 2, and the one that helped me learn Java the most is a book called Java 2: The Complete Reference (3^rd Edition) by Patrick Naughton and Herbert Schildt, a book that considers Java Applets to be the greatest human achievement, maybe second to fire if you are feeling generous.

Hey, wasn't this about "good code"?

Shit, I forgot... Basically, what I am trying to say is that because of how AI makes you "dumb", it makes your code worse as well. When you use AI to write code, you don't truly know what it is writing. When you use AI to learn how to write code, you don't truly learn.

So... what IS good code?

Good code is code that is transparent. Good code is code that is written with intent, by a human and for humans.

This doesn't mean never using code others wrote. That'd be extremely dumb. Rather, what I mean is that you should understand what it is that you are using. You should be able to explain the inner workings of the tool you are using. If you can't, that is completely fine, but it should signal to you that you need to learn more, and learn properly.

In the end, good code is about having fun if it is for a hobby, but it should also be about knowing what you write. That little difference takes your code from "good" to "amazing".