Engineering video
1:23 · 22 August 2025
Learning niche languages like Cobol or Rust might yield higher short-term compensation due to lower AI replacement risk currently, but mainstream languages like JavaScript or TypeScript offer much better long-term job security because of steady demand.
Hear it at 0:00 →LLMs will eventually learn niche languages as private codebase data becomes available, meaning the best defense against AI and job loss is mastering fundamentals in one language rather than learning many superficially.
Hear it at 0:38 →learning niche programming language like Cobol, Rust, Haskell, because AI is not very good in those languages and it will not replace today. While the reality is that niche languages have niche opportunities, so it's not like you have thousands of Cobol jobs. It's actually very few, so yes, maybe you make a bit more money on the short term. But in the long run, if you're unemployed for let's say six months, you've lost all that benefits.
Whereas if you know a mainstream language, like JavaScript or TypeScript, sure, maybe you're not able to demand like such a big compensation as someone that's more specialized, but you have the job security. Because there's always someone that needs somebody doing TypeScript or JavaScript. Remember that if it's written in a computer, you can write something on paper and an LLM can learn it. So, the reason why we don't have good LLMs for Cobol or Haskell or any of those languages, is because they weren't trained on that data set.
But as soon as we get that data set, as soon as they get into private code bases and they actually learn, you will have specialized LLMs for Cobol and for all these niche languages, and they'll do a lot of work there. It's not like you can gatekeep DLMs from learning it. And also remember that the biggest problem in this right now, it's a lot of developers don't have strong fundamentals, is not that they don't know enough languages, but they don't know a language well enough to be really expert there.
Become an expert at one thing first before you learn five new ones.
