It was only a few years ago that snooty media personalities were constantly echoing President Joe Biden’s advice to unemployed workers: “Learn to code”. Then, of course, the media hit hard times and the advice was then being snarkily offered to newly unemployed media folks. But what about the (few) who actually did “learn to code”, only to be swept away again as the clankers surged in to eliminate a lot of basic coding jobs?
How to Understand What AI Just Did to People Who Took Joe Biden’s Advice and Learned to Code.
A simple, concrete example.
Oddly enough, I have a bachelor’s degree in Computer Science. This means I know 7 algorithms for sorting a list into alphabetical order. I understand the tradeoffs between their execution time, code complexity, and memory demand. I learned the specialized lingo for describing execution time.
The algorithms are surprisingly complex and subtle. I spent months learning to code them.
Now that hard-won knowledge has been replaced by, “Claude, write a module to sort this list. Optimize for execution time.”
Millions of good people just lost their professions and must now invest in a new one.
Right now, knowing how to sort a list probably gives me a small advantage when I code with AI. But I will soon lose even that tiny return on my investment, as AI improves.
Certainly AI will create some new opportunities, probably a lot of them.
But count your blessings, if you did not spend years learning to code like I did.
And:
Here is the counterpoint: Learning to code gave coders an advantage when relearning to code with AI.
That advantage is their ticket to a seat in the new AI world.
The big question now is how many seats exist.
To which ESR responded:
Your position is reasonable, but wrong.
Having learned to code is still valuable in the new world of AI, not because you’re wrong about coding itself having become disposable, but because of the capabilities and mindset you developed while learning to code, some of which are difficult to learn in any other way.
You didn’t become a professional programmer. But I’m willing to bet that your intuition about how to design software is far better because you wrestled with code. And that is *not* a skill that LLMs are replacing — ignore the noisy hype about this.
I’m also willing to bet that some of what you learned as a programmer in training translated into problem-decomposition skills that have served you well as an economist.
If one is not a complete dullard (and you are certainly not a complete dullard) learning to code teaches not just craft skills but a mindset — a set of heuristics for carving reality at its joints. There are other ways to get this — I think for example of Richard Feynman who got there by thinking very hard about physics. And it is not guaranteed that every programmer will develop this right mindset.
But many of us do. And most of the other ways to develop it seem also to produce it only as a side effect, but less reliably than learning to code does.
So don’t write off learning to code. Maybe someday we’ll develop educational methods that can teach those higher-level skills more directly. That would be an excellent thing, if it’s possible. But until it gets here, learning to code will still have value that is not easy to duplicate in any other way.
